perfect waker for array::Merge

[yoshuawuyts]

Ref #21.

Speeds up array::merge by copying over the implementation from vec::merge. We can probably speed this up even further by creating a non-allocating variant of WakerList - we may need to for #68 anyway. But for now a 2-20x speedup seems like plenty, and we can switch out the impl later without any issue anyway.

❯ critcmp main patch  -f array::merge
group                main                                   patch
-----                ----                                   -----
array::merge 10      1.64      4.2±0.26µs        ? ?/sec    1.00      2.6±0.09µs        ? ?/sec
array::merge 100     9.35   376.3±25.42µs        ? ?/sec    1.00     40.3±1.95µs        ? ?/sec
array::merge 1000    21.88    40.0±1.72ms        ? ?/sec    1.00  1827.5±46.92µs        ? ?/sec

[eholk]

Looks good to me!

It'd be nice if we could factor out this code and the vec code and make it independent of the underlying container, but I think that's a problem for later.