Sergey "Shnatsel" Davidoff

According to the [profile](https://share.firefox.dev/3NdHKiD), this did actually parallelize things, but now it requires Huffman decoding to complete before Rayon kicks in. ~~The performance results are a toss-up: on some images...

Wait, I have actually measured it incorrectly. I need to turn off all the OTHER parallelization that Rayon does to make an apples-to-apples comparison. It might very well be that...

Ah, turning off parallel color conversion also gated on `rayon` feature makes this PR 25% slower even on its previously "good" cases, and 30% slower on the large CMYK test...

FYI `cargo afl build` actually compiles with optimizations but with debug assertions and overflow checks enabled, so it's much faster than true debug mode.

I believe the latter will, yes. I'm not 100% sure, test it if you need to rely on that behavior.

It would be nice if someone could re-test the performance now that `png` crate has switched from `inflate` to `miniz_oxide` in v0.16.5 Speaking of Rust PNG implementations, there are also...

Address Sanitizer and/or Thread Sanitizer could be used to pinpoint the memory safety issue, see https://github.com/japaric/rust-san I'm not sure if they work on Windows, though.

`zip` crate has recently switched to miniz_oxide as the compression backend. @mjc has anything changed for your program? Is the issue still reproducible?

AFAIK libraries cannot have their own allocators - there should be just one global allocator in the entire binary (which is why it's called "global"). So you could use this...

At a glance, most integration issues between Rodio and Symphonia have been fixed (#379, #401). #405 still outstanding, but the current mp3 adapter doesn't expose the total duration either, so...