fix(openapi-parser): strip a leading UTF-8 BOM before JSON.parse in normalize - #10792

Problem

@scalar/openapi-parser maintained its own normalize(), separate from the one in @scalar/json-magic that #9107 fixed. JSON saved with a leading UTF-8 BOM was rejected, so an otherwise valid API description could fail in validate(), dereference(), or upgrade().

For example, save this document as UTF-8 with a BOM:

{
  "openapi": "3.1.0",
  "info": { "title": "Example", "version": "1.0.0" },
  "paths": {}
}

Previously, validate(await readFile('openapi.json', 'utf8')) rejected the document. It now parses and validates successfully.

Solution

Import and re-export @scalar/json-magic/helpers/normalize directly, removing the parser's duplicate normalizer entirely. This keeps BOM handling and JSON/YAML parsing in one place. Objects and filesystems retain their identity.

The exported helper now has the shared signature (content: unknown) => unknown. TypeScript callers must narrow its result before accessing properties.

Normalization also adopts the shared helper's handling of malformed JSON arrays: inputs such as [{foo: "bar"}] return undefined instead of falling back to YAML parsing.

Ticket

See #9097