Preserve version order when legacy prerelease counters roll over - #500

Version.parse("1.2.3-rc9").next_version("prerelease") returns 1.2.3-rc10, which is older under SemVer's lexical prerelease ordering. That contradicts next_version's ordering guarantee and can make an automated release increment move backwards.

Check the candidate in the same-token legacy-counter branch. If it would not be newer, advance the patch while retaining the incremented legacy counter: 1.2.3-rc9 becomes 1.2.4-rc10. Ordinary increments, dotted numeric counters and already-forward leading-zero transitions retain their existing patch. Build metadata is cleared, subclasses are preserved and the input stays unchanged.

Fixes #499. This applies the patch-advance precedent from #498 to the separate legacy return path. It deliberately updates test_issue_484_rc9_to_rc10, whose old expectation pinned the backwards result, and documents that behavior in the method and news fragment. Open #494/#496 address long-identifier conversion limits and are outside this patch.

Validation on Windows:

• Original implementation with the new regressions: 8 failed, 5 controls passed.
• Original full suite: 498 passed, 1 skipped.
• Fixed full suite, including doctests: 511 passed, 1 skipped on CPython 3.12.14 and 3.13.5; 100% statement and branch coverage. The skip concerns the absent optional native parser.
• Ruff 0.15.9 reports the same 159 existing findings on both trees, with no new diagnostics. The new test file passes format check and git diff --check passes. The repository-wide lint is not claimed clean.

This Windows Git checkout materializes two pre-existing test symlinks as target-name text. Validation ran in a separate source copy with those two files resolved to the exact documented source bytes; neither workaround is included in the PR. Other platforms/interpreters and the native backend have not been exercised locally.

Found by code inspection and prepared with OpenAI Codex assistance. No production incident or independent human review is claimed.