Backport PR #31423: PERF: postpone imports in Index constructor (#31424)

Co-authored-by: Joris Van den Bossche <jorisvandenbossche@gmail.com>

Author: meeseeksmachine

pandas/core/indexes/base.py

@@ -278,10 +278,6 @@ def __new__(
     ) -> "Index":
 
         from pandas.core.indexes.range import RangeIndex
-        from pandas import PeriodIndex, DatetimeIndex, TimedeltaIndex
-        from pandas.core.indexes.numeric import Float64Index, Int64Index, UInt64Index
-        from pandas.core.indexes.interval import IntervalIndex
-        from pandas.core.indexes.category import CategoricalIndex
 
         name = maybe_extract_name(name, data, cls)
 
@@ -297,10 +293,16 @@ def __new__(
 
         # categorical
         elif is_categorical_dtype(data) or is_categorical_dtype(dtype):
+            # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
+            from pandas.core.indexes.category import CategoricalIndex
+
             return CategoricalIndex(data, dtype=dtype, copy=copy, name=name, **kwargs)
 
         # interval
         elif is_interval_dtype(data) or is_interval_dtype(dtype):
+            # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
+            from pandas.core.indexes.interval import IntervalIndex
+
             closed = kwargs.pop("closed", None)
             if is_dtype_equal(_o_dtype, dtype):
                 return IntervalIndex(
@@ -315,6 +317,9 @@ def __new__(
             or is_datetime64_any_dtype(dtype)
             or "tz" in kwargs
         ):
+            # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
+            from pandas import DatetimeIndex
+
             if is_dtype_equal(_o_dtype, dtype):
                 # GH#23524 passing `dtype=object` to DatetimeIndex is invalid,
                 #  will raise in the where `data` is already tz-aware.  So
@@ -329,6 +334,9 @@ def __new__(
                 return DatetimeIndex(data, copy=copy, name=name, dtype=dtype, **kwargs)
 
         elif is_timedelta64_dtype(data) or is_timedelta64_dtype(dtype):
+            # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
+            from pandas import TimedeltaIndex
+
             if is_dtype_equal(_o_dtype, dtype):
                 # Note we can pass copy=False because the .astype below
                 #  will always make a copy
@@ -339,6 +347,9 @@ def __new__(
                 return TimedeltaIndex(data, copy=copy, name=name, dtype=dtype, **kwargs)
 
         elif is_period_dtype(data) or is_period_dtype(dtype):
+            # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
+            from pandas import PeriodIndex
+
             if is_dtype_equal(_o_dtype, dtype):
                 return PeriodIndex(data, copy=False, name=name, **kwargs).astype(object)
             return PeriodIndex(data, dtype=dtype, copy=copy, name=name, **kwargs)
@@ -358,6 +369,13 @@ def __new__(
 
         # index-like
         elif isinstance(data, (np.ndarray, Index, ABCSeries)):
+            # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
+            from pandas.core.indexes.numeric import (
+                Float64Index,
+                Int64Index,
+                UInt64Index,
+            )
+
             if dtype is not None:
                 # we need to avoid having numpy coerce
                 # things that look like ints/floats to ints unless