revlog: break up compression of large deltas
Python's zlib apparently makes an internal copy of strings passed to
compress(). To avoid this, compress strings 1M at a time, then join
them at the end if the result would be smaller than the original.
For initial commits of large but compressible files, this cuts peak
memory usage nearly in half.
# should print nothing
# should fail
abort: path 'b/x' is inside repo 'b'
abort: path 'b/x' is inside repo 'b'
# should fail
abort: path 'b/x' is inside repo 'b'
# should arguably print nothing
# should fail
abort: path 'b/a' is inside repo 'b'