Fix heapq and _heapq stubs

[jonathandung]

Closes #15666 and #15668.

[github-actions]

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

spark (https://github.com/apache/spark)
+ python/pyspark/shuffle.py:561: error: Argument "key" to "merge" has incompatible type "Callable[[V], Any] | None"; expected "Callable[[V], SupportsDunderLT[Any] | SupportsDunderGT[Any]]"  [arg-type]
+ python/pyspark/core/rdd.py:2575: error: Argument "key" to "nlargest" has incompatible type "Callable[[T], S] | None"; expected "Callable[[T], SupportsDunderLT[Any] | SupportsDunderGT[Any]]"  [arg-type]
+ python/pyspark/core/rdd.py:2578: error: Argument "key" to "nlargest" has incompatible type "Callable[[T], S] | None"; expected "Callable[[T], SupportsDunderLT[Any] | SupportsDunderGT[Any]]"  [arg-type]
+ python/pyspark/core/rdd.py:2635: error: Argument 3 to "nsmallest" has incompatible type "Callable[[T], S] | None"; expected "Callable[[T], SupportsDunderLT[Any] | SupportsDunderGT[Any]]"  [arg-type]
+ python/pyspark/core/rdd.py:2637: error: Argument 3 to "nsmallest" has incompatible type "Callable[[T], S] | None"; expected "Callable[[T], SupportsDunderLT[Any] | SupportsDunderGT[Any]]"  [arg-type]

[JelleZijlstra]

The new pyspark errors in the mypy-primer output are unfortunate. Can we avoid them?

[jonathandung]

They appear to be fixable by introducing a new type variable, but I don't know if it's worth the trouble to open an issue over there.