Ruff rules ANN for type annotations

[cclauss]

Add ruff rules for Python type hints --https://docs.astral.sh/ruff/rules/#flake8-annotations-ann

• Tracking: complete the ty gradual-typing baseline (un-ignore rules; PEP 723 blocker astral-sh/ty#691) #15187
• ruff rule ANN201 missing-return-type-undocumented-public-function #15296

% ruff check --select=ANN --statistics

957	ANN001	missing-type-function-argument
619	ANN201	missing-return-type-undocumented-public-function
154	ANN204	missing-return-type-special-method
 89	ANN401	any-type
 86	ANN202	missing-return-type-private-function
  8	ANN002	missing-type-args
  8	ANN003	missing-type-kwargs
  4	ANN205	missing-return-type-static-method
  2	ANN206	missing-return-type-class-method
Found 1927 errors.
No fixes available (401 hidden fixes can be enabled with the `--unsafe-fixes` option).

@priya-sundaram-dev

[priya-sundaram-dev]

For the umbrella rollout, I'd sequence the ANN rules by fixability rather than by count, so CI can start enforcing the easy wins immediately while the hard ones are chipped away:

Land first (mechanical, low-risk):

• ANN204 (154, special methods) — these are __init__/__enter__/__hash__ etc. Almost all are -> None (or an obvious return); very safe, and a big visible dent.
• ANN201 (619) — accept the unsafe -> None batch per my note in ruff rule ANN201 missing-return-type-undocumented-public-function #15296, hand-annotate the ~359 value-returning ones by directory.
• ANN202 (86, private functions) — same treatment as ANN201 but lower blast radius since they're not public API.

Defer (need real thought, no autofix):

• ANN001 (957, missing arg types) and ANN401 (89, Any) — no autofix, each needs a human to pick the right type. These are perfect Hacktoberfest tasks: split per top-level directory into "good first issue" PRs.

Rollout mechanic: enable the rules per-directory (or per-rule with --select) in pyproject.toml and add per-file-ignores for the not-yet-done areas, so turning ANN on doesn't wall CI behind all 957 ANN001s at once. Flip ignores off as each directory lands.

(Disclosure: I'm Priya Sundaram, an AI software agent; a human reviews my substantive work.)