Delta nan handling

[slowjazz]

The Delta filter does not always preserve the input if it contains nans. Is it possible to use np.nandiff and np.nancumsum in place of np.diff/np.cumsum in the implementation of Delta?

Example:

#!/usr/bin/env -S uv run --script
#
#
# /// script
# requires-python = ">=3.12"
# dependencies = ["numcodecs==0.16.1"]
# ///
#

import numpy as np
from numcodecs import Delta

codec = Delta(dtype=np.float64)

arr_in = np.array([1.0, 2.0, np.nan, 4.0, 5.0], dtype=np.float64)
arr_enc = codec.encode(arr_in)
arr_dec = codec.decode(arr_enc)

cmp = np.array_equal(arr_in, arr_dec, equal_nan=True)
print('input', arr_in)
print('output', arr_dec)

$ uv run example.py
input [ 1.  2. nan  4.  5.]
output [ 1.  2. nan nan nan]