Python: Split type checkers by target (pyright source, 5 checkers on tests/samples)

[eavanvalkenburg]

Motivation and Context

Following the "too many type checkers"
approach, this reworks our typing setup so that each target is checked by the
right tool(s):

• The source ran two strict checkers (mypy + pyright), doubling the
  annotation burden and producing checker-specific friction on internal code.
• Tests and samples were under-checked, so type problems in the way customers
  actually call the public API only surfaced downstream.

The goal: make source development easier (one source checker) while giving
customers more confidence that running any of the five common type checkers
against their own code will surface fewer surprises.

Description

• Source: Pyright (strict) is now the sole source-code type checker.
  MyPy is removed from source; its [tool.mypy] block becomes a relaxed
  profile used only for tests/samples.
• Tests: checked by all five checkers — pyright (relaxed), mypy, pyrefly,
  ty, and zuban.
• Samples: checked by pyright, pyrefly, and ty (mypy/zuban can't resolve
  script-style sample layouts).
• The relaxed/basic profiles intentionally disable noisy rules (private-access,
  not-required TypedDict access, untyped test bodies) so test/sample authors
  aren't forced into ugly over-annotation. Narrow, rule-specific ignores
  (# pyright: ignore[rule], # type: ignore[code]) are used only where a
  checker is pedantic.
• Added pyrightconfig.tests.json; bumped sample pyright configs to basic.
• Added pyrefly.toml, pyrefly.samples.toml, ty.samples.toml.
• Unified test/sample typing onto the same parallel fan-out used by source
  pyright (run_command_items in scripts/task_runner.py) for consistent
  execution.
• Made version-conditional imports symmetric: the # type: ignore is now kept
  or dropped on both branches, so results match across interpreter versions
  (local vs CI) instead of only matching the current venv.
• Updated SKILL.md, DEV_SETUP.md, and CODING_STANDARD.md for the five
  gating checkers and pyright over source + tests + samples.

Contribution Checklist

• The code builds clean without any errors or warnings
• The PR follows the Contribution Guidelines
• All unit tests pass, and I have added new tests where possible
• Is this a breaking change? If yes, add "[BREAKING]" prefix to the title of the PR.

[Copilot]

Copilot wasn't able to review this pull request because it exceeds the maximum number of files (300). Try reducing the number of changed files and requesting a review from Copilot again.