perf: avoid extra parsing and codegen while enabling minify

[hyf0]

What problem does this feature solve?

Now: Module AST -> per_codegen -> Chunk string -> parse to chunk AST -> swc minify -> Chunk AST codegen After: Module AST -> merge as a Chunk AST -> swc minify -> Chunk AST codegen in parallel

We could save the extra time for codegen and parsing.

What does the proposed API of configuration look like?

No response