Copy to buffer breaks if dtype changes.

[WenzDaniel]

Describe the bug
Not necessarily a bug, but more of a missing feature? If the dtype used in copy_to_buffer is reduced by a field, but using the same function name it fails. This is problematic for example when adding removing the data_top field in straxen peaks.

See screenshot below for an example.

[dachengx]

With MWE copied from @WenzDaniel linked image:

import numpy as np
import strax

a = np.ones (1, strax.peak_dtype(digitize_top=True))
b = np.zeros(1, strax.peak_dtype(digitize_top=True))
strax.copy_to_buffer(a, b, '_my_copy_function_name')
a = np.ones(1, strax.peak_dtype(digitize_top=False))
b = np.zeros(1, strax.peak_dtype(digitize_top=False))
strax.copy_to_buffer(a, b, '_my_copy_function_name')

There is no error anymore, thanks to #785.