perf: avoid closures that cause allocations

[costela]

This is a light improvement on allocations and performance:

pkg: github.com/exaring/hoglet
cpu: 12th Gen Intel(R) Core(TM) i7-1250U
                           │  old.bench   │              new.bench              │
                           │    sec/op    │    sec/op     vs base               │
Hoglet_Do_EWMA-12            409.7n ± 12%   376.9n ± 17%       ~ (p=0.075 n=10)
Hoglet_Do_SlidingWindow-12   463.2n ±  2%   453.3n ±  7%       ~ (p=0.149 n=10)
geomean                      435.6n         413.4n        -5.10%

                           │ old.bench  │             new.bench              │
                           │    B/op    │    B/op     vs base                │
Hoglet_Do_EWMA-12            240.0 ± 0%   208.0 ± 0%  -13.33% (p=0.000 n=10)
Hoglet_Do_SlidingWindow-12   240.0 ± 0%   208.0 ± 0%  -13.33% (p=0.000 n=10)
geomean                      240.0        208.0       -13.33%

                           │ old.bench  │             new.bench              │
                           │ allocs/op  │ allocs/op   vs base                │
Hoglet_Do_EWMA-12            6.000 ± 0%   5.000 ± 0%  -16.67% (p=0.000 n=10)
Hoglet_Do_SlidingWindow-12   6.000 ± 0%   5.000 ± 0%  -16.67% (p=0.000 n=10)
geomean                      6.000        5.000       -16.67%

Still some work to do. We should be able to get to 1-2 allocs per call.
I do not think zero allocations are possible as long as we have the asynchronous reactions to context deadlines/cancelations, since that needs sharing across goroutines.

[obitech]

LGTM