Implement NONCOPY for the diploid case

[szhan]

Fix #58

[szhan]

The tests test_ts_simple_n8_high_recomb in TestNonTreeViterbiDiploid seem stuck. The CI runs started two hours ago, and they still haven't finished.

[szhan]

The naive implementations of the diploid Viterbi algorithm are taking too long when the ref. panel size gets too large when including ancestral haplotypes.

[szhan]

A workaround is to only run the naive implementations of diploid Viterbi when not including ancestors in the ref. panels.

[szhan]

Alternatively, we could check the size of the ref. panel when including ancestors. If it exceeds a certain threshold, say n = 100, then don't run the naive implementations of diploid Viterbi.

[szhan]

For now, when the ref. panel size exceeds 100, the naive implementations of diploid Viterbi are not run. There is probably a better way to do this.