Preserve RegularEnum before enum optimization

When a namespace comes before an enum it merges with (namespace Foo { ... } enum Foo { A = 1 }) and optimizeEnums is enabled, this branch records Foo as only BlockScopedVariable. apply_pending_symbol_flags runs in enter_program before the traversal reaches TypeScriptEnum::enter_expression, and that inliner gates on flags.contains(SymbolFlags::RegularEnum), so later Foo.A references stop being inlined even though they were before this change and the option asks for regular enum inlining. Delay dropping RegularEnum until after enum inlining or keep enough enum state for that pass.

Useful? React with 👍 / 👎.

This finding is real, it seems break following code

namespace Foo {
  export const x = 1;
}
enum Foo {
  A = 1,
}
Foo.A; // before is `1`, now is `Foo.A`

Preserve enum flags until enum optimization runs

When a value namespace is the first declaration in a merge and an enum/const enum with the same name appears later, this overwrites the merged symbol flags with only BlockScopedVariable during TypeScriptNamespace::enter_program, before TypeScriptEnum traverses the program. With optimizeEnums or optimizeConstEnums enabled, TypeScriptEnum::resolve_enum_member checks symbol_flags for RegularEnum/ConstEnum, so cases like namespace E { export const x = 1 } enum E { A = 1 } console.log(E.A) stop inlining E.A even though the enum data is still available. Please keep the enum flags available until the enum pass has run, or avoid clobbering non-namespace flags here.

Useful? React with 👍 / 👎.