[codex] phase1: bridge root excision (no metasignature redesign)

[darinkishore]

Summary

This PR executes Phase 1: Bridge Root Excision by removing the legacy bridge crates from the workspace while intentionally preserving current runtime contracts.

Problem and Root Cause

The repository still carried the legacy bridge runtime and derive macro crates even after facet-based typed parity work was completed. Those crates increased maintenance overhead and obscured the intended architecture by keeping dead roots alive.

What Changed

• Deleted crates/baml-bridge entirely.
• Deleted crates/baml-bridge-derive entirely.
• Updated docs to mark bridge-root removal as the active baseline and to note that Phase 2 will remove compat-trait coupling and redesign signature/optimizer APIs.
• Kept runtime contracts intact in this phase (bamltype::compat, LegacySignature, LegacyPredict, MetaSignature, optimizer APIs).
• Kept typed/legacy wiring behavior unchanged for this cut.

User Impact

• Expected breaking change: direct consumers of removed bridge crates will fail to compile until migrated.
• Existing in-repo typed and legacy paths continue to pass existing tests in this phase.

Verification

All requested gates were executed successfully:

• rg -n "baml-bridge|baml_bridge|baml-bridge-derive" /Users/darin/src/personal/DSRs --glob '!target/**'
• cargo check --workspace
• cargo fmt --all
• cargo clippy --workspace --all-targets -- -D warnings
• cargo test -p bamltype --tests
• cargo test -p dsrs_macros --tests
• cargo test -p dspy-rs --tests
• cargo test -p dspy-rs --test test_bamltype_attr_contract
• cargo test -p dspy-rs --test test_bamltype_docs_contract
• cargo test -p dsrs_macros --test signature_derive
• cargo test --workspace

Follow-up (Phase 2)

• Remove remaining compat-trait coupling from typed paths.
• Redesign signature/optimizer APIs to be facet-native.