Made bench_heavy_awk.ct and bench_heavy_bash.ct identical except for
@awk decorator. Fixed compound assignment with method calls that was
causing different results.

New features:
- .map() / .filter() for arrays with lambda support
- @validate decorator for argument validation
- fs.open() / f.read() / f.write() / f.close() file handles
- with X in Y { } context manager for auto-closing resources

All features work in both Bash and @awk codegen.
Added 18 new pytest tests for the new functionality.
Updated documentation (LANGUAGE_SPEC.md, README.md, README_ru.md).

- @test("description") decorator marks functions as tests
- assert(condition, message) works in both bash and @awk functions
- content test command runs all @test functions
- Colored output with timing, pass/fail status
- DCE skips @test functions in normal compilation mode
- Supports combining @test with other decorators (@awk, @retry, etc.)

- Track which classes are instantiated (new X() or X())
- Track transitive dependencies via class fields
- Skip generating unused classes with comment marker

Example: if class A is used but class B is not, B is skipped:
  # DCE: skipped unused class B

- Add dce.py: UsageAnalyzer that scans AST for used stdlib categories
- Modify stdlib.py: emit_stdlib() now accepts used_categories parameter
- Modify codegen.py: analyze programs before generating stdlib

Categories: core, object, http, fs, json, logger, string, array,
dict, regex, math, time, awk, exception, args, misc

Results:
- Simple "print(Hello)" program: 200+ lines -> 35 lines
- Complex programs: minimal change (use most categories anyway)

- Add estrlist.ct: string list operations with @awk acceleration
- Add estrlist_noawk.ct: pure bash version for comparison
- Add benchmark.sh: performance comparison script
- Add BENCHMARK.md: results showing @awk gives 93x speedup on union

Codegen fixes:
- Fix time.now() and time.ms() calls in dispatch_codegen
- Fix math.* namespace calls
- Previous fixes for class field assignment and nested constructors