(re-factor) gc: remove redundant since_sweep field

[topolarity]

This field is at all times 0 or identical to allocd, so this change removes it in preference of gc_num.allocd.

Hopefully this makes the GC metrics a tiny bit easier to interpret for newcomers (like myself).

[d-netto]

My concern is whether people could be using gc_num.since_sweep. Probably unlikely since it's GC internal, but still.

[topolarity]

My concern is whether people could be using gc_num.since_sweep. Probably unlikely since it's GC internal, but still.

I suppose if we want to avoid the breaking change we can provide a wrapper in timing.jl to fill in the removed field?

[KristofferC]

JuliaHub shows no uses in packages at least.

[d-netto]

Seems like we use since_sweep internally at RAI but are in the process of updating the GC metrics we export, which would eliminate the use of such metric among other things.

I believe removing this metric would be fine from our perspective.

Tagging @NHDaly for more input on that.