phase A of workspace consolidation. collapses the `anno` facade crate
(which only re-exported `anno-lib`) and renames the `anno-lib` package
to `anno`. the lib target was already named `anno`, so `use anno::...`
keeps working; users on `anno = "0.7"` get a normal upgrade with no
source change.

publish chain drops from 7 to 6 crates: anno-core, anno-metrics,
anno-graph, anno, anno-eval, anno-cli. phase B (folding anno-core /
anno-metrics / anno-graph into `anno`) is planned for a future release.

skips the partial 0.7.0 publish: anno-core 0.7.0 and anno-metrics 0.7.0
are on crates.io but the rest of the chain failed mid-publish on a
trusted-publisher misconfiguration. rather than re-running 0.7.0 for
content about to be folded, rolls forward to 0.8.0 directly.

other:
- root Cargo.toml becomes a virtual workspace (no [package] section).
- delete src/lib.rs and src/main.rs (facade only).
- consumers updated: anno-cli, anno-eval drop `package = "anno-lib"`.
- publish.yml drops the redundant facade publish step.
- CITATION.cff bumped 0.7.0 -> 0.8.0, date 2026-04-26.
- doc cleanups: PUBLISH_STATUS rewritten to reflect current state;
  CONTRACT removes facade language; QUICKSTART bumps anno = "0.8" and
  fixes Entity field-access example to use accessor methods; deprecation
  notes added to anno-core and anno-metrics READMEs.

rationale and ecosystem evidence in .claude/reports/research-2026-04-26.md
(tokio's 2017 reform, BurntSushi's split criterion, 14-repo survey).