Opt-in FunctionDefinition collection tree node (#3926)

[RonnyPfannschmidt]

Closes #3926.

Promotes the internal FunctionDefinition from a transient parametrization
helper to a real collector node in the collection tree, behind an opt-in ini
option.

collect_function_definition (tri-state string, default hidden)

Mode	Tree	Function-level markers
hidden (default)	flat, definition kept out of the tree	on the Function (legacy)
pedantic	definition node inserted	scoped to the definition; invocation owns only its callspec marks
messy	definition node inserted	transferred back onto each invocation (legacy layout), + header warning

messy is a migration stopgap for code that is not yet prepared for the
definition scope — it reproduces the old duplicated-marker layout so bad code
keeps working. The target is pedantic.

What changes

• FunctionDefinition becomes a PyCollector (no longer a Function/Item), so
  it is collected rather than run. Its collect() runs pytest_generate_tests
  and yields the invocations.
• Invocation node ids stay flat and identical across all modes
  (test_mod.py::test_x[0]), so -k, caching, --lf, xdist and reporting are
  unaffected.
• obj/instance resolution and getmodpath transparently skip an interposed
  definition node.
• Marker scoping is correct in every mode (iter_markers yields each mark once).

The default (hidden) path is behavior-preserving; the whole test suite passes
unchanged.

Known warts / follow-up TODOs

These are intentionally left as follow-ups (inline TODO(#3926) where relevant),
to keep this PR focused — happy to split into a tracking issue:

1. Item-scoped getfixtureinfo on a collector. The fixture closure must be
   computed before the invocations exist, so FunctionDefinition._generate_functions
   calls the item-scoped getfixtureinfo on the collector via a cast. Needs a
   node-level entry point for closure computation. (TODO(#3926) inline.)
2. messy mode is temporary. It exists only to unblock migration and should
   be removed once consumers move to the definition scope. Marked clearly in the
   docs and header warning.
3. Collector/item node-id collision. For non-parametrized tests the definition
   node and its sole Function child share a node id. Accepted for now; the
   structured-nodeids work will give these a better data structure.
4. originalname could now become a read-only property returning the
   definition's name (see the existing note in Function.__init__).

🤖 Generated with Claude Code

[RonnyPfannschmidt]

Heads-up on a defect in this PR as it currently stands: selecting a test by node id is broken in the tree modes.

$ cat pytest.ini
[pytest]
collect_function_definition = pedantic

$ pytest -q "test_x.py::test_a[1]"
ERROR: not found: test_x.py::test_a
(no match in any of [<Module test_x.py>])

Works under hidden, fails under both pedantic and messy. So the PR description's

Invocation node ids stay flat and identical across all modes, so selection, caching and reporting are unaffected.

holds for the node id strings, but not for selection: Session._collect_and_match_nodes walks the tree comparing one name part per level, and the last part is matched as node.name == part + parametrization. With the definition interposed, alpha[1] is compared against the definition's bare name alpha and never matches, so it never descends to the items. -k and full-path invocation are fine — only ::-addressing of a specific test is affected.

The minimal fix is to let the matcher look through an ItemDefinition for the last part (match the bare name, then re-match the same part one level down against the items). I have that plus a regression matrix (6 selector shapes × 3 modes) in #14805, which is stacked on this branch — happy to move it down into this PR instead if you'd rather this one be self-contained.

Worth noting the existing comment right at that spot already anticipates this:

# A non-parameterized arg matches all parametrizations (if any).
# TODO: Remove the hacky split once the collection structure
# contains parametrization.
is_match = node.name.split("[")[0] == matchparts[0]

This PR is exactly what makes the collection structure contain the parametrization, so that split("[") hack is now removable in principle — though doing it properly probably wants the structured NodeId/ParamId from #14758 rather than more string surgery. Will follow up on that.