Fix XZ decompression

[WallaceIT]

Add a decompression buffer and some logic to fix XZ decompression.

[WallaceIT]

Hi! I tested this with a real-world BMAP file, but more testing is advised before merging.

While working on this, I noticed a potential flaw in the write approach (not only for xz but for the entire application): if an area of the source image is very large, this can lead to an excessive memory allocation for the write buffer. Of course, the meaning of "very large" here depends on the system it is used on. A better approah would probably be to split decompression (or read, in the case of no compresssion) in smaller chunks and then calculate the checksum in an incremental way.

[embetrix]

Fix #1

[embetrix]

@WallaceIT : thank you for the fix, I did some tests on my stm32mp1 with 512M and very large chuncks ~100MB, it went okay.
you are right that will be a good approach, more over I would to use libarchive instead of having both liblzma and zlib.