Sized fastpath

[arielb1]

r? @eddyb

The change to trans::common::type_is_sized is because we currently abort, rather than return random results, on overflow.

[eddyb]

AFAICT there is no chance of leakage of cached results that depend on the parameter environment, though there might be other gotchas I'm not aware of.
LGTM but I'd really have r? @nikomatsakis on this.

[arielb1]

I am using the same caching logic as before.

[eddyb]

You're totally right, my bad. @bors r+

[bors]

📌 Commit 595409d has been approved by eddyb

[arielb1]

Performance numbers:

Before:
581.72user 4.75system 7:42.74elapsed 126%CPU (0avgtext+0avgdata 1176224maxresident)k
llvm took 359.183

After:
573.72user 4.49system 7:34.47elapsed 127%CPU (0avgtext+0avgdata 1166328maxresident)k
llvm took 356.015

[eddyb]

@arielb1 is that saying it took... longer? @bors r-
EDIT: I've reverted the r-, but do note what @rust-highfive did: it matched "longer?" as "r?".
cc @nrc

[arielb1]

I mixed them up

[eddyb]

@bors r+

[bors]

📌 Commit 39e6855 has been approved by eddyb

[bors]

⌛ Testing commit 39e6855 with merge 75fb009...

[bors]

☀️ Test successful - auto-linux-32-nopt-t, auto-linux-32-opt, auto-linux-64-nopt-t, auto-linux-64-opt, auto-linux-64-x-android-t, auto-mac-32-opt, auto-mac-64-nopt-t, auto-mac-64-opt, auto-win-gnu-32-nopt-t, auto-win-gnu-32-opt, auto-win-gnu-64-nopt-t, auto-win-gnu-64-opt