fix(mobile): use circular particle heading moments - #735

…es (#768)

Annotates the tools, demo, bin, models, ets and robot packages and holds them to disallow_untyped_defs. BaseETS is generic over its element type (BaseETS[ET], BaseETS[ET2]), replacing the self-type overloads. 68 unused pyright ignores removed, module/class name clashes made explicit. Left on the mypy to-do list: tools/trajectory.py (open #736), mobile/ParticleFilter.py (open #735), the two unused DHFactor modules and a stale test in tools/urdf/tests. Behaviour unchanged except on paths that already raised.

… cross-references, zero warnings (#833)

* docs(ets): ETS gets its own section, with 3D and 2D pages, tables and ET constructors

ets.rst replaces arm_ets.rst: corrected line references and the fkine
return type, and ETS is now a top-level item. ets_3d.rst and ets_2d.rst
document ET/ETS and ET2/ETS2 with Information, Modification and Kinematics
tables, the constructors, joint descriptor strings with examples, and the
three inverse kinematics mechanisms.

Fixes docstring examples that failed when built (ET2/ETS2 copied from the
3D case, split, swap, merge), docutils warnings, and adds missing
docstrings for plot, teach, ETS2.jacob0, qlim and s.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: remove the page-level Jesse Haviland credit, he is first author in the footer

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs(ets): overview table lists the page, the sequence class and the element class

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* feat(ets): fkine and eval take no argument for a sequence without joints

q is now optional on ETS and ETS2 fkine and eval, and means an empty joint
vector. A sequence with joints still raises a ValueError that names the
lengths when q is missing or too short; tests pin both behaviours.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs(ets): say angle or distance, not amount

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs(ets): public class names, constructor arguments, joint descriptor rules as text, IK examples

Documents ET, ETS, ET2 and ETS2 under their public names, drops the module
prefix from signatures (add_module_names = False), lists the constructor
arguments, moves the joint descriptor rules out of the BaseET class box,
adds fixjoint to the Modification tables, and rewrites the opening examples
and the IK section.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: fix typos in docstrings and pages, and the warnings the docs build printed

Typos found with codespell in rendered docstrings and docs/source (ParticleFilter,
DHFactor, Dynamixel and ROS left alone). A docstring with an escaped asterisk
is now raw. The quintic example no longer calls plot() in a runblock, where
there is no display. Robot.manipulability(method='asada') uses eigvalsh,
since eig returned a complex array for a matrix symmetric to rounding error,
which printed a ComplexWarning.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: correct the introduction, IK, mobile and blocks pages after a read-through

Stale line references and a named-tuple claim in the introduction, a nonexistent
ETSLink and manipm, a misplaced History section, URDF models described as DH,
the planner tables, a wrong seg-fault claim for the C++ IK solvers, dead
placeholders (:py:class:`Class`), and the broken bdsim logo URL.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ik): a q0 of the wrong length is an error, in the C++ and the Python solvers

The C++ solvers (ik_LM, ik_NR, ik_GN) ignored a q0 of the wrong length and
started from a random vector, so the caller got a solution for a different
start than the one they asked for. The Python solvers raised an unhelpful
NumPy broadcast error. Both now raise ValueError naming the length expected,
through one helper, check_q0.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ik): a mask that is not 6 elements is an error

The C++ solvers read six elements from the mask, so a shorter one was read
past its end; the Python solvers failed later with a matrix shape error.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: describe the q0 and mask checks of the C++ IK solvers; explain eigvalsh

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* fix(ets): swap accepts transforms that commute, merge rejects a joint, ET2.s is a 3-vector

swap only accepted two transforms of the same kind although it is documented
for commutative ones: it now accepts translations along different axes and a
rotation with a translation along its axis, and names the pair when it refuses.
merge raised TypeError for a joint and a constant, now ValueError. ET2.s
returned a 4-vector, a copy of the 3D case, now the 3-vector twist. Two
seealso references to SerialLink.jtraj now point at jtraj.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: resolve the cross-references that the strict build reports

Fix at the source the references that named a short or a wrong target (ETS, IKSolver.solve,
plot_bg, spatialmath.SE3, Link.*, DHRobot.*), turn names that were never
objects into literals, index the methods of the IK solver classes so they can
be linked, document URDFRobot and NoPathError, and list in nitpick_ignore_regex
the few kinds that can never resolve. The strict build (sphinx-build -n) went
from 1650 warnings to 5, three of them in ParticleFilter, left for #735.

Also the partial_fkine0 example prints the shape of the result, not the matrix.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

* docs: two more cross-reference fixes (hessian0, a missing colon in a :math: role)

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

---------

Co-authored-by: Claude Sonnet 5.5 <noreply@anthropic.com>