Use dynamic huffman tables to reduse the crate size.

[RazrFalcon]

This commit makes this crate 4x times smaller without performance changes.

Sizes from the image-png crate:

Before:

% cargo bloat --release --example pngcheck --filter inflate
 File  .text     Size Name
10.4%  70.7% 506.2KiB [1307 Others]
 4.2%  28.7% 205.3KiB inflate::InflateStream::next_state
 0.0%   0.3%   2.0KiB inflate::DynHuffman16::new
 0.0%   0.1%   1.0KiB inflate::InflateStream::run_len_dist
 0.0%   0.1%     508B inflate::DynHuffman16::read
 0.0%   0.1%     475B inflate::InflateStream::update
 0.0%   0.0%     201B inflate::InflateStream::from_zlib
14.7% 100.0% 715.7KiB .text section size, the file size is 4.8MiB

After:

% cargo bloat --release --example pngcheck --filter inflate
 File  .text     Size Name
10.7%  91.0% 506.2KiB [1307 Others]
 1.0%   8.2%  45.9KiB inflate::InflateStream::next_state
 0.0%   0.4%   2.0KiB inflate::DynHuffman16::new
 0.0%   0.2%   1.0KiB inflate::InflateStream::run_len_dist
 0.0%   0.1%     508B inflate::DynHuffman16::read
 0.0%   0.1%     475B inflate::InflateStream::update
 0.0%   0.0%     201B inflate::InflateStream::from_zlib
11.8% 100.0% 556.3KiB .text section size, the file size is 4.6MiB

Benchmark results using libflate/flate_bench:

Before:
-                 inflate: elapsed=1.903052s, size=288632530
After:
-                 inflate: elapsed=1.869871s, size=288632530

[RazrFalcon]

Looks like the build is failed due to ? operator. Should I use try!?

[eddyb]

@RazrFalcon Yeah, I think so.

[eddyb]

Actually, you should .unwrap(), sorry - these can't ever fail, they're effectively constants.

[RazrFalcon]

But why it returns 'Result' then?

[eddyb]

I meant these calls can't ever fail because they have constant arguments - for other inputs, those functions may return Err.

[RazrFalcon]

@eddyb done.