Update to upstream reed-solomon-erasure

[gnunicorn]

In an attempt to investigate whether we are effected by the memory leak known in reed-solomon-erasure, this updates the dependency – switching from our fork to the upstream version, which is just ahead of ours (includes all our PR).

Unfortunately I do not know enough about this particular part of the code base to understand whether this might have unintended side-effects. Please think about that hard when reviewing.

[burdges]

I never noticed this runs inside the runtime.

[rphmeier]

@burdges I don't believe it does - the polkadot-erasure-coding crate is not linked to the runtime in my knowledge.

[rphmeier]

We should not be affected by this memory leak regardless, as that code path isn't hit without parachains and we don't use a decoder more than once before discarding.

[burdges]

We'd improve performance by using the decoder more than once however, so.. Is the erasure coding an appreciable performance hit? If so, I'll take a look at whether we want systemic coding, etc. and the available optimizations.

[rphmeier]

Yup, although this isn't something we've encountered yet. I'm expecting erasure-coding to be a major bottleneck.