Use IPC module init extended data (dp_data lifetime and interim heap
sizes) to determine DP module heap size when available. Add Kconfig
option SOF_USERSPACE_DP_DEFAULT_HEAP_SIZE (default 20480) as fallback
when extended init data is not present or does not provide heap sizes.

Sanity-check the requested sizes (reject values above 64 MB or below
the dp_heap_user prefix) and log the allocated heap size.

Also pass ext_init through module_adapter_mem_alloc() to
module_adapter_dp_heap_new() and fix a minor comment typo.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Use IPC module init extended data (dp_data stack_bytes) to set the DP
processing thread stack size when available. Fall back to
TASK_DP_STACK_SIZE when ext init data is not present or does not
provide a stack size.

Add Kconfig option ZEPHYR_DP_SCHEDULER_MIN_STACK_SIZE (default 2048)
to enforce a minimum stack size regardless of what the IPC payload
requests.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Name DP threads using k_thread_name_set() with the component ID
in hex. This makes it easier to identify DP threads in debug tools
and Zephyr shell thread listings.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

The widget specific stack requirement works in such a way that all
components in a pipeline have their stack requirements, and the
pipeline stack size will be the highest of those values.

I am not sure if this is actually the best method. I am not sure what
factors affect the required stack size for a pipeline, but for now
this will have to do.

So 4k is not enough for all pipelines, and these generic numbers are
here to guarantee that everything works until we have fine tuned
values for all the modules. So increase the number to 8k to guarantee
functionality.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Fine tune SRC DP memory requirements to values that should satisfy all
use-cases, without using too much memory.

For now the biggest memory consumers, the SRAM copy of the filter
coefficients, and the filter memory, are not allocated in init phase,
but in prepare. That is why the interim requirement is so high and
lifetime requirement so low. This can be changed but it needs changes
in SRC processing module code.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Ring buffers at LL/DP boundaries are created during pipeline_connect
and persist until module free. They are never freed and reallocated
across prepare/reset cycles, so they belong in the lifetime partition
rather than interim.

This allows DP modules to use a smaller interim partition since the
ring buffer struct and data buffer no longer consume interim space.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Refactor the vregion memory layout to use a single contiguous buffer
instead of two separately page-aligned partitions. The vregion struct
is placed at the base, followed by lifetime allocations growing upward.
The interim k_heap is created lazily on the first interim allocation
from whatever space remains after lifetime allocations.

This eliminates the rigid partition boundary that previously wasted
memory when lifetime usage was smaller or larger than pre-configured.
The interim heap creation is deferred until actually needed, at which
point the lifetime region is sealed and any further lifetime allocation
requests are redirected to the interim heap with a warning.

Key changes:
- vregion_create(): Sum lifetime+interim into one buffer, no k_heap
  init at creation time
- New interim_heap_init(): Called lazily, page-aligns interim start,
  logs lifetime used and interim available at INFO level
- lifetime_alloc(): Falls back to interim with warning if called after
  interim is initialized
- interim_alloc(): Triggers interim_heap_init() on first call
- vregion_free(): Guards interim pointer range check with
  interim_initialized flag

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Add a third memory type that uses lifetime before the interim heap is
initialized and interim after, without triggering interim heap creation
or logging a warning. This is suitable for small metadata allocations
that do not care which partition they reside in.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Object pool metadata does not need to reside in a specific partition.
Use INDIFFERENT so pool blocks go to lifetime before the interim heap
is created and to interim afterwards, without triggering premature
interim heap initialization.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Add a vregion_type field to struct mod_alloc_ctx to track the current
vregion allocation mode for each module. Initialize it to
VREGION_MEM_TYPE_LIFETIME during module_adapter_mem_alloc() so that
all allocations during module creation and init use the lifetime
partition. Switch it to VREGION_MEM_TYPE_INTERIM at the start of
module_adapter_prepare() so that runtime allocations use the interim
partition.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Use vregion_type from strcut mod_alloc_ctx, when vregion is available,
for buffer allocations.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Get mod_alloc() vregions vregion_mem_type from struct mod_alloc_ctx.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Use vregion_type from struct mod_alloc_ctx to control which vregion
memory type fast_get() uses for allocations.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>

Refactor SRC and SRC-lite so that filter stage setup and delay line
allocation happen during the module init callback instead of prepare.
This ensures the bulk of SRC memory allocation occurs while the vregion
allocator is still in its lifetime phase, before the interim heap is
created. The allocations then persist across prepare/reset cycles
without needing to be re-allocated each time.

A new setup_stages() callback is added to struct comp_data, set by each
variant (src.c, src_lite.c) to point at its own coefficient tables.
The common src_allocate_delay_lines() is factored out of the old prepare
path into src_common.c.

For IPC4, src_init_stages() calls setup_stages() and
src_allocate_delay_lines() at init time. The prepare path
(src_prepare_do) only validates rates and sets downstream params.

For IPC3, src_init_stages() is a no-op and src_prepare_do() retains
the original behavior of doing full setup at prepare time, since IPC3
cannot be tested at this time.

Signed-off-by: Jyri Sarha <jyri.sarha@linux.intel.com>