Add Scala support to repository analysis

[mageshwaranr]

What

Adds Scala support, addresses #105.

How

New analyzers/scala.py following the same tree-sitter pattern as the Kotlin/Ruby
analyzers:

• Components: classes, case classes, traits, objects, package objects, Scala 3
  enums, methods, and top-level functions. Uses a two-level component_type /
  node_type split — component_type is the coarse type the pipeline gates leaf
  selection on (traits → interface, objects → class, so both stay leaf-eligible and
  count toward the issue-C/C++ repository collapses to 5 leaf nodes during dependency analysis. Silent fallback to whole-repository mode #75 OOP heuristic), while node_type/display_name preserve
  the faithful Scala construct (trait, object, enum). This avoids widening the
  shared OOP_TYPES set, which would change documentation output for PHP/Kotlin/C++ as
  a side effect — a decision left for its own change.
• Companion objects: get a $-suffixed component id (buffer.scala::Buffer vs.
  buffer.scala::Buffer$), mirroring Scala's own JVM encoding of module classes, so a
  class and its same-named companion never displace each other.
• Relationships: inheritance, trait mixins, constructor-parameter and field types
  (including Scala 3 parameterized traits), instantiation, and method calls — resolving
  intra-file targets against a same-file symbol table and leaving external targets as
  unresolved logical names. A curated primitive/stdlib-call exclusion set (mirroring the
  Kotlin/PHP precedent) keeps Int, List, .map, .size, etc. out of the graph.
• sbt build files: build.sbt and project/*.sbt classify as build
  manifest/artifact files (alongside the existing Maven/Gradle handling) rather than
  being parsed as application source; .sbt is deliberately never added to
  CODE_EXTENSIONS.

Registration follows the Ruby/Kotlin precedent end-to-end: both language whitelists in
analysis_service.py, the dispatcher branch + lang-scala cytoscape class in
call_graph_analyzer.py, .sc mapping and Scala function patterns in patterns.py,
module-path stripping in ast_parser.py, the source-extension set in artifact.py, the
code-fence language map in prompt_template.py, CLI repo validation and language
detection, and MCP incremental-change detection.

Dependency note: pins tree-sitter-scala==0.23.4, not the newer 0.26.2. Verified
empirically that 0.24.0+ compiles against grammar ABI 15, which raises
Incompatible Language version 15 under this repo's tree-sitter==0.23.2 core pin —
declared package metadata (tree-sitter~=0.22) doesn't reflect the actual compiled ABI.
0.23.4 is the last ABI-14 release and is node-for-node identical to 0.26.2 for every
construct the analyzer reads (verified by diffing field-annotated parse trees), so no
tree-sitter core bump is needed for this change.

Tests

tests/test_scala_analyzer.py (11 tests): component extraction (classes, traits,
objects, companion $ suffix, docstrings, parameters including curried parameter
groups), relationship extraction (resolved/unresolved inheritance and mixins,
constructor-parameter edges including Scala 3 parameterized traits, intra-file calls,
unresolved external calls across all three fallback paths, stdlib-noise exclusion),
Scala 3 enum/significant-indentation syntax, sbt artifact classification, .sc
extension end-to-end, empty-file handling, and a DependencyParser end-to-end test
proving the whitelist/dispatcher wiring and cross-file inheritance resolution.

Full existing suite (110 tests) passes with no regressions in the Python, Kotlin, PHP,
Ruby, or artifact analyzer tests. ruff check / ruff format --check clean on every
touched file.

🤖 Generated with Claude Code

[anhnh2002]

Thank you very much @mageshwaranr. Merged!