perf(diffctx): eliminate duplicate get_ignore_specs call (33% speedup)

Pass combined_spec through DiscoveryContext instead of recomputing
in DefaultDiscovery. Profiling showed get_ignore_specs was called
twice, consuming 6.8s of 10.9s total (63%). Now called once.
Measured: 10.87s → 7.26s on treemapper repo.

Author: nikolay-e

src/treemapper/diffctx/pipeline.py

@@ -244,6 +244,7 @@ def build_diff_context(
             diff_text=diff_text,
             expansion_concepts=frozenset(expansion_concepts),
             file_cache=file_cache,
+            combined_spec=combined_spec,
         )
         discovered_files = _create_discovery(config).discover(discovery_ctx)
         discovered_files = [_normalize_path(p, root_dir) for p in discovered_files]

src/treemapper/diffctx/scoring.py

@@ -3,6 +3,7 @@
 from abc import ABC, abstractmethod
 from dataclasses import dataclass
 from pathlib import Path
+from typing import Any
 
 from .graph import Graph
 from .types import DiffHunk, Fragment, FragmentId
@@ -16,6 +17,7 @@ class DiscoveryContext:
     diff_text: str
     expansion_concepts: frozenset[str]
     file_cache: dict[Path, str] | None = None
+    combined_spec: Any | None = None
 
     def read_file(self, path: Path) -> str | None:
         if self.file_cache is not None and path in self.file_cache:
@@ -42,7 +44,7 @@ def discover(self, ctx: DiscoveryContext) -> list[Path]:
 
         from ..ignore import get_ignore_specs
 
-        combined_spec = get_ignore_specs(ctx.root_dir, None, False, None)
+        combined_spec = ctx.combined_spec or get_ignore_specs(ctx.root_dir, None, False, None)
         expanded = _expand_universe_by_rare_identifiers(
             ctx.root_dir,
             ctx.expansion_concepts,