check the size of seqs before trying to decode them

[jmesmon]

Fixes #41

Not sure if this style OOM issue affects other parts of the decoder.

[jmesmon]

It looks like at the least read_map() also needs help in this area.

[jmesmon]

It isn't clear to me that this is really the right solution: it presumes that size_of is the same size as the representation of T when bincoded (which isn't generally correct).

It is probably sufficient to assume that for this check, but I'm a bit concerned about potential corner cases where the size_of is significantly larger than bincoded T.

Perhaps the real solution here is to account separately for bytes read vs bytes allocated?

Maybe this is good enough for now?

[TyOverby]

Oh wow. This is a pretty big issue; the kind that I had hoped to fix with SizeLimit in the first place.

I think this could be solved in a few other ways, but probably not as easily as this.

Because in-memory size is just as important as wire-size, I think I should rewrite the docs saying that it can fail with SizeLimit on either.

[TyOverby]

Fixing read_map is implemented in 628b989