Cache type_object_type()

[ilevkivskyi]

This gives almost 4% performance boost (Python 3.12, compiled). Unfortunately, using this in fine-grained mode is tricky, essentially I have three options:

• Use some horrible hacks to invalidate cache when needed
• Add (expensive) class target dependency from __init__/__new__
• Only allow constructor caching during initial load, but disable it in fine-grained increments

I decided to choose the last option. I think it has the best balance complexity/benefits.

[ilevkivskyi]

Non-trivial results in mypy_primer are surprising. I know there is a bug in type_object_type() that when it is called before the constructor is processed we may get wrong result (e.g. because it is decorated) instead of deferring the current node. But I thought the way I implemented caching should not be affected by this bug. I will take a look today or tomorrow.

[github-actions]

Diff from mypy_primer, showing the effect of this PR on open source code:

spark (https://github.com/apache/spark)
- python/pyspark/pandas/indexes/multi.py:1180: error: Argument 1 to "_index_fields_for_union_like" of "Index" has incompatible type "DataFrame[Any] | Series[Any] | Index | list[Any]"; expected "Index"  [arg-type]
+ python/pyspark/pandas/indexes/multi.py:1177: error: Redundant cast to "MultiIndex"  [redundant-cast]
+ python/pyspark/pandas/indexes/category.py:249: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:273: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:295: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:340: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:370: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:431: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:484: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/category.py:558: error: Incompatible return value type (got "Index", expected "CategoricalIndex | None")  [return-value]
+ python/pyspark/pandas/indexes/base.py:2021: error: Invalid index type "int | Any | tuple[Any, ...] | list[int | Any | tuple[Any, ...]]" for "list[Any | tuple[Any, ...]]"; expected type "SupportsIndex"  [index]
+ python/pyspark/pandas/indexes/base.py:2091: error: Argument 1 to "from_tuples" of "MultiIndex" has incompatible type "Index"; expected "list[tuple[Any, ...]]"  [arg-type]

[ilevkivskyi]

Although remaining errors are kind of correct, they are sill weird. It looks like they appear because previously

class C
    @no_type_check
    def __new__(cls): ...

resulted in Any as the type of C class object. But somehow it is now C in case of an import cycle. FWIW I can't reproduce this in a simple test.