Expand static-analysis normalization coverage for PHPStan pseudo-types

[Copilot]

The static-analysis path handled basic refined types, but did not sufficiently validate PHPStan-specific pseudo-types against native/manual-safe stub types. This adds targeted coverage and tightens normalization so PHPStan call map types compare cleanly with stub signatures.

• Normalization updates

  • Extend TypeNormalizer to collapse additional PHPStan-only syntax into comparable native types:
    • callable signatures with declared return types
    • resource (closed)
    • nullable shorthand handling at the start of a type
  • Keep comparisons aligned with existing native/manual-safe output rather than preserving PHPStan-specific detail.

• Regression coverage

  • Add a dedicated static-analysis fixture covering representative PHPStan types:
    • list<T>
    • class-string<T>
    • array shapes
    • pure-callable(...)
    • iterable<K, V>
    • int-mask<...>
    • resource (closed)
  • Add a CLI regression test asserting the static_analysis command treats these inputs as equivalent to native/phpDoc stub types.

• Docs

  • Clarify in the README that the static-analysis flow normalizes PHPStan-only pseudo-types back to comparable native types before reporting mismatches.

Example of the newly covered input shape:

return [
    'phpstan_type_match' => [
        'list<string>',
        'value' => 'non-empty-string',
        'callback' => 'pure-callable(int):literal-string',
        'class' => 'class-string<stdClass>',
        'items' => 'iterable<int,positive-int>',
        'map' => 'array{foo:int,bar?:string}',
        'payload' => 'object{foo:int}',
        'mask' => 'int-mask<1, 2, 4>',
        'stream' => 'resource (closed)',
    ],
];