feat(pixi-build): support package.run-constraints

[hunger]

Add [package.run-constraints] (and [package.target.<sel>.run-constraints]) to express conda run_constrained metadata on packages built with pixi. Constraints apply to packages in the run environment only when they are pulled in by another dependency — they never cause installation on their own, mirroring conda's existing semantics.

The new field is threaded end-to-end:

• pixi_manifest: new SpecType::RunConstraints; TOML parsing for the [package.run-constraints] and target-scoped variants.
• pixi_build_types: new Target.run_constraints field on the protocol type, included in is_empty/Hash and round-tripped via serde.
• pixi_build_type_conversions: to_target_v1 propagates the bucket.
• pixi_build_backend: PackageSpecDependencies and the Targets / Dependencies trait surfaces gain the new field; from_targets_v1_to_conditional_requirements emits constraints into Requirements.run_constraints for both default and per-selector targets. The wire protocol's CondaBuildV1Params.run_constraints was already in place; the dispatcher already routes them to rattler-build.
• schema/model.py + regenerated schema.json / pyproject schemas so IDE autocomplete and validators see the new field.
• docs: new section in docs/reference/pixi_manifest.md, a brief entry in docs/build/dependency_types.md, and a snippet block in the package-manifest TOML fixture.

Tests cover each conversion layer: TOML parsing, target-specific routing, the pixi-build preview gate on source specs in run-constraints, the to_target_v1 and From<&Target> conversions, from_targets_v1_to_conditional_requirements populating both default and platform-conditional entries, and a regression test for Target::is_empty so a future copy-paste bug can't silently drop a target whose only payload is run-constraints.

Closes #4871
Closes #5418

How Has This Been Tested?

New unit tests. I think this covers the functionality quite well. I also tried a manual test project, which worked fine for me.

AI Disclosure

• This PR contains AI-generated content.
  • I have tested any AI-generated content in my PR.
  • I take responsibility for any AI-generated content in my PR.

Tools: Claude

Checklist:

• I have performed a self-review of my own code
• I have commented my code, particularly in hard-to-understand areas
• I have made corresponding changes to the documentation
• I have added sufficient tests to cover my changes.
• I have verified that changes that would impact the JSON schema have been made in schema/model.py.

[ruben-arts]

I tested it locally and it works with a pixi-build-rust backend.

[baszalmstra]

I have only one comment! Otherwise look good!