Chunked phase-retrieval reconstruction incorrectly recover slices on borders

[nikitinvv]

Example:
If chunk size 32 then slices 0..4 and 28..31 shows no effect of the Paganin filtering.

For correct filtering one should use the chunks of sizes at least 48 (40 for ichunk=0 and ichunk=end-1), i.e. 0..40, 24..72, and cut them after filtering to get 0..32, 32..64, etc.

[nikitinvv]

Solved in this commit

[nikitinvv]

after/before:

[decarlof]

@nikitinvv shall I add this in the readme file or start a sphynx/readthedoc? Perhaps the last is better so we can also add better documentation for other functionalities (see hardening added by @aps-7bm etc.)

[nikitinvv]

@decarlof, agree. it would be good to write about that in readthedocs. I think this problem may also occur in beam hardening