Inline Default impls and trivial instance/device wrapper methods

[Ralith]

I did not inline load, any of the wrappers that invoke read_into_*_vector helpers, or (for now) extensions methods. In theory this should be a nice clean overhead reduction at runtime. Data on buildtime impact would be great!

[MarijnS95]

So it seems my random yet completely unrepresentative workstation machine is now used for benchmarking 😂

Before:

$ hyperfine -p 'cargo clean -p ash' 'cargo b -p ash'
Benchmark 1: cargo b -p ash
  Time (mean ± σ):      5.841 s ±  0.031 s    [User: 6.671 s, System: 0.373 s]
  Range (min … max):    5.786 s …  5.898 s    10 runs

After:

$ hyperfine -p 'cargo clean -p ash' 'cargo b -p ash'
Benchmark 1: cargo b -p ash
  Time (mean ± σ):      5.724 s ±  0.026 s    [User: 6.435 s, System: 0.364 s]
  Range (min … max):    5.691 s …  5.782 s    10 runs

~120ms.

[Ralith]

Looks like a slight win on my end too. Nice!