Fix MultiIndex sel with tuple-valued levels (GH#11341)

[C1-BA-B1-F3]

Problem

When a MultiIndex level contains tuple-valued entries (e.g., (1,1)), selecting with a nested tuple key like ((1,1), 2) incorrectly preserves the dimension instead of collapsing it to a scalar result.

Example from Issue #11341

import xarray as xr
import numpy as np
import pandas as pd

# Create MultiIndex where first level contains tuples
nested_level_0 = pd.Index([(1, 1), (1, 1), (2, 2), (3, 3)], name="a")
nested_level_1 = pd.Index([1, 2, 10, 20], name="b")
nested_mi = pd.MultiIndex.from_arrays([nested_level_0, nested_level_1])

nested = xr.DataArray(
    np.arange(4),
    dims=("index",),
    coords={"index": nested_mi},
)

# This should collapse the dimension (shape=()) but returns shape=(1,)
result = nested.sel(index=((1, 1), 2))

Root Cause

The _is_nested_tuple() function was checking for tuple in addition to list and slice:

def _is_nested_tuple(possible_tuple):
    return isinstance(possible_tuple, tuple) and any(
        isinstance(value, tuple | list | slice) for value in possible_tuple
    )

This caused tuple-valued keys like ((1,1), 2) to be incorrectly identified as nested selection tuples (containing multiple selection values), when they are actually single keys with a tuple-valued first level.

Fix

Remove tuple from the isinstance check so only list and slice are treated as indicators of nested selections:

def _is_nested_tuple(possible_tuple):
    return isinstance(possible_tuple, tuple) and any(
        isinstance(value, list | slice) for value in possible_tuple
    )

This correctly distinguishes:

• (1, 2) → single key (scalar values) → use get_loc()
• ((1,1), 2) → single key (tuple value in first level) → use get_loc()
• ([1,2], 3) → nested selection (list) → use get_locs()
• (slice(1,2), 3) → nested selection (slice) → use get_locs()

Tests

Added regression test test_sel_nested_tuple_key that verifies:

1. MultiIndex with tuple-valued levels can be selected with nested tuple keys
2. The dimension is correctly collapsed (scalar result, not length-1)

Closes #11341