Decompress after reading in blocks

[tlnagy]

As identified in #17, a key area of weakness is my strategy for inflating compressed data. Currently, I inflate while reading the data, which is likely slower than reading a contiguous block followed by in-memory inflation. Plus, I should consider spinning out (some of) the compression code into separate package for reusability sake.

[tlnagy]

Plus, I should consider spinning out (some of) the compression code into separate package for reusability sake.

The compression code is being moved over to https://github.com/tlnagy/Balloons.jl

[chrstphrbrns]

#127 implements chunked input

[tlnagy]

Agreed, this issue is mostly superfluous in v0.8.0+