Implement OptiType pre-filtering step

[tavinathanson]

After #423 is done, we can consider using a better aligner that doesn't suffer from the following:

You can use any read mapper to do this step, although we suggest you use RazerS3. Its only drawback is that due to way RazerS3 was designed, it loads all reads into memory, which could be a problem on older, low-memory computing nodes.

Suggested by @ihodes in https://github.com/hammerlab/biokepi/issues/419#issuecomment-279821401.

[ihodes]

Discussion here: https://github.com/hammerlab/biokepi/issues/423

[tavinathanson]

Per https://github.com/hammerlab/biokepi/issues/419#issuecomment-280073430, it may not be worth it to add this with razerS3 vs. an aligner that won't require 208GB+ of memory (or no possible GCP machine type for high enough depth BAMs).

[tavinathanson]

@armish and I agreed that the razers3 pre-filtering step (vs. using a different aligner) doesn't seem particularly high priority, since (a) there's a good chance it wouldn't impact results (and it shouldn't) and (b) it probably wouldn't fix OOM issues.