red-knot: move symbol lookups in symbol.rs

[dhruvmanila]

Summary

This PR does the following:

• Moves the following from types.rs in symbol.rs:
  • symbol
  • global_symbol
  • imported_symbol
  • symbol_from_bindings
  • symbol_from_declarations
  • SymbolAndQualifiers
  • SymbolFromDeclarationsResult
• Moves the following from stdlib.rs in symbol.rs and removes stdlib.rs:
  • known_module_symbol
  • builtins_symbol
  • typing_symbol (only for tests)
  • typing_extensions_symbol
  • builtins_module_scope
  • core_module_scope
• Add symbol_from_bindings_impl and symbol_from_declarations_impl to keep RequiresExplicitReExport an implementation detail
• Make declaration_type a pub(crate) as it's required in symbol_from_declarations (binding_type is already pub(crate)

The main motivation is to keep the implementation details private and only expose an ergonomic API which uses sane defaults for various scenario to avoid any mistakes from the caller. Refer to #16133 (comment), #16133 (comment) for details.

[AlexWaygood]

Nice! It's great to move these out of types.rs.

Bikeshedding: I'm wondering if we should put these in a new submodule (lookup.rs?) rather than symbol.rs. The APIs in symbol.rs are currently quite abstract and don't have much to do with the business of implementing Python's lookup semantics correctly; these APIs we're moving here feel a little different from that. I don't have a strong opinion, however; I'm happy to go with this

We could also add some more doc-comments to the module_type_symbol function, as I mentioned in #16133 (comment)

[carljm]

Looks good to me! I agree with all of Alex's comments.

[dhruvmanila]

Bikeshedding: I'm wondering if we should put these in a new submodule (lookup.rs?) rather than symbol.rs. The APIs in symbol.rs are currently quite abstract and don't have much to do with the business of implementing Python's lookup semantics correctly; these APIs we're moving here feel a little different from that. I don't have a strong opinion, however; I'm happy to go with this

We added a LookupResult in symbol.rs in #16160 which makes me wonder if we should rename symbol.rs to lookup.rs instead.

But, I also see this comment on LookupResult:

ruff/crates/red_knot_python_semantic/src/symbol.rs

Lines 176 to 177 in 89cefbe

	/// Note that this type is exactly isomorphic to [`Symbol`].
	/// In the future, we could possibly consider removing `Symbol` and using this type everywhere instead.

which makes me hesitant to do it now and can be done easily in the future as well.