Tuples are inferred as containing literal value types despite explicit annotation with general type

```py
t1: Tuple[int, ...] = (1, 2, 3)
t2: Tuple[int, ...] = (1, 2, 4)

t1 >= t2
```

`Operator ">=" not supported for types "tuple[Literal[1], Literal[2], Literal[3]]" and "tuple[Literal[1], Literal[2], Literal[4]]"`

`Pylance language server 2026.2.1 (pyright version 1.1.408, commit 138238b4)`

This is a regression of #1170. Specifically the case mentioned in https://github.com/microsoft/pyright/issues/1170#issuecomment-727048387. @erictraut [said](https://github.com/microsoft/pyright/issues/1170#issuecomment-727212010) that he added a test case for this to the test suite, you should check that it is still there.

Start by reproducing the reported comparison with the two annotated tuples, then inspect the existing test case referenced from issue #1170 in the test suite. Trace the tuple type inference involved in that case; done means explicit Tuple[int, ...] annotations no longer produce literal-element comparison errors and the regression test passes.

Written by the indexing model from the issue text.