big-endianness for size

[facboy]

it would be great if you could specify in the method calls (loads and dumps) whether the size header was little endian or big endian.

[steeve]

The size is always stored in little endian, to guarantee interop between archs.
See https://github.com/steeve/python-lz4/blob/master/src/python-lz4.c#L44

[facboy]

actually i'm reading data from a different library (hadoop as it happens), and its block compression uses a completely different scheme. basically it has the original size as a big-endian int32, followed by the compressed size as a big-endian int32, and then the compressed data. so a simple option wouldn't have worked in any case. maybe the ability to specify the size and then treat the data as merely the compressed stream?