M8-03: Add memory/performance budgets

[DrunkOnJava]

Created from the roadmap audit in TODO.md.

Source: https://github.com/DrunkOnJava/rvt-rs/blob/main/TODO.md#L967
Tracking commit: 1ec8105

Scope

Labels: priority:P1, type:test, area:ci

• Define budget targets for:
  • open
  • summarize
  • schema parse
  • element decode
  • IFC export
  • viewer parse/render
• Track against small, medium, and large corpus files.

Acceptance criteria:

• Benchmarks emit machine-readable results.
• CI or nightly flags regressions above threshold.

Definition of done

• The TODO acceptance criteria for this section are implemented or explicitly superseded by a linked decision.
• Tests, fixtures, or documentation are added at the level appropriate to the change risk.
• User-facing behavior and limitations remain honest in README/docs/viewer messaging.
• Relevant CI checks pass before the issue is closed.

[DrunkOnJava]

Completed in eeb37d2 with CI follow-up 4b286b7.

Implemented the runtime budget gate with machine-readable budgets/results, small+medium corpus enforcement, a scheduled/manual GitHub Actions workflow, and benchmark documentation.

Verification:

• Local: python3 -m py_compile tools/perf_budget.py, JSON validation, YAML parse, cargo fmt --all -- --check, cargo clippy --all-targets -- -D warnings, wasm clippy, cargo check --manifest-path rvt-py/Cargo.toml, cargo test --all-targets, cargo bench --no-run, doctests, rustdoc warnings, git diff --check, and stub-pattern scan all passed.
• Runtime gate: tools/perf_budget.py --enforce --require-category small --require-category medium passed locally.
• GitHub: manual Performance Budgets workflow run 24928792018 passed on 4b286b7; all standard commit checks for 4b286b7 are green.