feat(control_plane): Phase 2 step 1 -- merge schema.rs onto asap-ir

[zzylol]

Summary

Stacked on #392 (base branch set accordingly -- merge #392 first, then
this). First step of Phase 2 (docs/migration-plan-backend-plan.md):
schema.rs.

Column/ColumnId/DataType/Schema/CseError/cse_reuse_is_legal
are now re-exported from asap_ir::intent_algebra::schema instead of
defined locally. Unlike AggIntent's merge in #392 (which needed a
boundary-conversion layer because AccuracyTarget's wire shape actually
changed), this one is a clean full swap -- a fresh diff showed asap_ir's
version is a purely additive, backward-compatible superset:

• Column gains table: Option<String> (SQL join qualifier) +
  Column::new()/with_table() constructors.
• Schema gains closed: bool (schema-on-read completeness flag) +
  column_id_qualified().
• DataType is byte-identical, no changes at all.

Both new fields are #[serde(default)], confirmed backward-compatible by
asap_ir's own tests (schema_closed_defaults_to_open_when_absent,
column_table_defaults_to_none_when_absent). That's what made a full
swap the right call here instead of #392's boundary-conversion approach
for Column/DataType -- and it let me delete the
to_asap_column/from_asap_column/to_asap_dtype/from_asap_dtype
helpers agg_intent.rs needed in #392, since there's only one
Column/DataType type now.

Blast radius was smaller than expected

The migration plan flagged ~38 Column{} literals as the risk here.
Turned out most of those grep hits were field references or doc comments,
not constructions -- only 15 real struct-literal sites across 9 files
needed the new field added (table: None / closed: false).

Verification

• cargo build --workspace -- clean
• cargo test -p control_plane --lib -- 820 passed, same 1 pre-existing
  failure as feat(control_plane): Phase 1b -- depend on asap-ir for real, not a copy #392 (confirmed unrelated)
• rustfmt applied

Remaining Phase 2 work

Per the migration plan: expr_ir.rs (new file, D2 decision), fresh diff
on query_expr.rs/relational.rs, fresh diff on
binder.rs/column_resolution.rs, cse.rs move to L4 (with the
regression pass that requires), lower.rs, then retargeting the
PromQL/SQL frontends.

Test plan

• cargo build --workspace
• cargo test -p control_plane --lib
• Merge feat(control_plane): Phase 1b -- depend on asap-ir for real, not a copy #392 first, then this

🤖 Generated with Claude Code