feat: rebase onto upstream main — port FTS5 BM25 search, Interface registration, embeddings, cross-repo infrastructure

Rebased our PR #162 features onto upstream's latest main (commit 1d30971)
which includes MinHash SIMILAR_TO edges, CBM_CACHE_DIR, and major refactoring.

Ported features (building clean on upstream's refactored codebase):

1. FTS5 BM25 search infrastructure:
   - Contentless FTS5 virtual table (nodes_fts) with camelCase token splitting
   - cbm_camel_split() SQLite function: updateCloudClient → 'update Cloud Client'
   - FTS5 backfill in both full pipeline and incremental pipeline
   - Incremental reindex now preserves FTS5 (was wiping to 0 rows)

2. Interface registration in symbol registry:
   - Added 'Interface' to label filter in process_def() (pass_definitions.c)
   - Added 'Interface' to label filter in register_and_link_def() (pass_parallel.c)
   - Fixes: C# class Foo : IBar now creates INHERITS → Interface edges

3. C# base_list extraction:
   - Added 'base_list' to fallback base_types[] in extract_base_classes()

4. Embeddings infrastructure (opt-in via CBM_EMBEDDING_URL):
   - embeddings table in SQLite schema
   - cbm_cosine_sim() SQLite function for vector search
   - embedding.c/h: HTTP client, text generation, RRF merge, pipeline integration
   - Auto-generates embeddings during indexing when configured

5. Cross-repo infrastructure:
   - cross_repo.c/h: unified _cross_repo.db builder, cross-repo search,
     channel matching, trace helper

Not yet ported (follow-up commits):
- MCP tool changes (search_graph query param, generate_embeddings tool,
  cross-repo tools, get_impact tool)
- Process detection (cbm_store_detect_processes)
- Channel detection (cbm_store_detect_channels)
- C# delegate event subscription (extract_calls.c)
- WRITES expansion (extract_semantic.c)

All upstream features preserved: MinHash SIMILAR_TO, pass_similarity,
CBM_CACHE_DIR, TS_FIELD() macro, extracted helpers.

Author: Your Name

Makefile.cbm

@@ -145,7 +145,7 @@ PREPROCESSOR_SRC = $(CBM_DIR)/preprocessor.cpp
 SQLITE_WRITER_SRC = $(CBM_DIR)/sqlite_writer.c
 
 # Store module (new)
-STORE_SRCS = src/store/store.c
+STORE_SRCS = src/store/store.c src/store/cross_repo.c
 
 # Cypher module (new)
 CYPHER_SRCS = src/cypher/cypher.c
@@ -186,7 +186,8 @@ PIPELINE_SRCS = \
     src/pipeline/pass_compile_commands.c \
     src/pipeline/pass_infrascan.c \
     src/pipeline/pass_k8s.c \
-    src/pipeline/pass_similarity.c
+    src/pipeline/pass_similarity.c \
+    src/pipeline/embedding.c
 
 # SimHash / MinHash module
 SIMHASH_SRCS = src/simhash/minhash.c

internal/cbm/extract_defs.c

@@ -955,6 +955,7 @@ static const char **extract_base_classes(CBMArena *a, TSNode node, const char *s
                                        "implements_clause",
                                        "argument_list",
                                        "inheritance_specifier",
+                                       "base_list", /* C# class Foo : IBar */
                                        NULL};
     return find_base_from_children(a, node, source, base_types);
 }