Correct calculation of dot_length in multi-channel case, fixing buffer overrun

Add getEffectiveRatio

Fix overrun

Add reset

Refactor e2e tests, add another one

Copy ctor; some test material

Some minor debug/reporting improvements

Add quality option etc to example resampler

Fix multi-channel support

Start on multi-channel support - not yet working

Ah, the meat of that last change wasn't ever going to work - the often-changing case requires the same p-multiple for the filter as the prototype, and I'm not sure how far it's wise to have the two methods use different filter lengths

Quality adjustments

Quality settings

Avoid reallocations when ratio changes in OftenChanging mode

Backed out changeset a78a79668677 - it is measurably slower on M1

More float buffers

Factor out common make_filter (interpolating the kaiser part, usually)

Fix failure to ensure odd-length filter

Make some use of bqvec functions

Pull out into .cpp file, tidy, wire up some of the options, add example utility