fix(workspace): unbreak main after #285/#291/#292 merge collision

[coseto6125]

Summary

main is currently red on 162c52db — `cargo check --workspace --all-targets --all-features` fails with 5 errors. The cause: #285 (T1-4/T1-5/T1-11), #291 (T4-7), and #292 (T7-2) all changed the same schema surface (`Node` struct, `Node.uid` type, `RawSchemaField` fields) but did not rebase against each other, so the merged state has call sites that don't match the merged-state types.

This PR is a forward-only reconciliation — no history rewrite, no force-push. Each fix brings a call site to the shape the original PR author would have written if they had rebased against the other two. End state is identical to a hypothetical "replay the PRs in correct dependency order" branch.

Changes (3 files, ~19 LOC)

• `post_process/schema_field_mirrors.rs:99-118`
  • Replace `format!()` + `string_pool.add()` UID with `uid::compute(NodeKind::SchemaField, &path_str, Some(owner_name), field_name)` — T1-5 canonical pattern (matches `resolution/builder.rs:372`).
  • Add now-required `content_hash: 0` (synthetic mirror node, no source span — per T7-2 "0 for synthetic nodes" doc convention).
  • Add now-required `owner_class: string_pool.add(owner_name)` (T1-11 rename isolation key — a `Foo.id` SchemaField correctly belongs to class `Foo`).
  • Side effect: drops one heap alloc per SchemaField mirror (no more `format!()` + pool dedup round-trip).
• `protobuf/parser.rs:101-110`
  • Replace `pool.add(&field_name)` and `pool.add(owner)` with `field_name.into_boxed_str()` and `Box::from(owner.as_str())` — T4-7's chosen Box form.
  • Drop the now-unused `pool` parameter from `extract_proto_fields` + the `StringPool::new()` at the call site (no external callers — verified).
• `python/parser.rs:17`
  • Add `use ecp_core::pool::StringPool;` — the T5-2 event-topic wire-up at line 1094 references `StringPool::new()` but the import was dropped somewhere during merges.

Audit scope (nothing deferred)

Searched the entire workspace for any related issues hiding behind the same merge collision:

• All `Node { ... }` literal sites (27 sites across builder.rs, coverage.rs, cypher/executor.rs, process_trace, leiden, test helpers) — all already include `content_hash` and `owner_class` correctly; only `schema_field_mirrors.rs` was missing.
• All `RawSchemaField { ... }` construction sites (protobuf + `schema_field/extract.rs`) — `schema_field/extract.rs` already uses `name.into()` / `owner.into()` (correct Box); only protobuf needed the migration.
• All callers of `extract_proto_fields` — none external (one doc-comment-only mention in `tests/protobuf_schema.rs`); signature change safe.
• All remaining `pool.add(...)` calls in post_process / protobuf / schema_field paths — they're all for `Edge.reason` or `Node.name` (both `StrRef`, correct usage), not stale `RawSchemaField` assignments.
• `uid::compute` callsite consistency — my new call matches the canonical `(kind, path, owner_class_or_empty, name)` ordering used at `resolution/builder.rs:372` and `commands/coverage.rs:556-661`.

Verification

• `cargo check --workspace --all-targets --all-features` — clean (was 5 errors before)
• `cargo clippy --workspace --all-targets --all-features -- -D warnings` — clean
• `cargo clippy --workspace --all-targets --all-features -- -D warnings -W clippy::all` — also clean (no pedantic regressions)
• `cargo test --workspace --no-fail-fast` — 2805 passed, 0 failed, 15 ignored
• `cargo test --test schema_field_mirror` — 6 passed (covers my schema_field_mirrors.rs edit)
• `cargo test --test protobuf_schema` — 13 passed (covers my protobuf parser edit)
• `cargo build -p egent-code-plexus --bin ecp --release` — 0 errors (323 crates)
• No `NodeKind` / `RelType` / `FrameworkId` enum changes (no rkyv schema disturbance)
• No `GRAPH_FORMAT_VERSION` bump (binary layout untouched)
• No hot-path edits (`pre_tool_use::handle`, `compute_hits`, `dispatch_by_mode` untouched)
• CI on push: Code Quality (Linting & Formatting) now SUCCESS (was the failing job on main)