LocusWalkerSpark drops intervals with no reads

[tomwhite]

This PR fixes #3823 opened by @LeeTL1220

There are several series of commits applied here:

1. The first set of commits rebase ll_CollectAllelicCountsSpark on master.
2. Then there's a commit (21e1dcf) to fix up some changes to make CollectAllelicCountsSpark work on master.
3. The next set of commits applies the changes from Use SparkFiles to localize reference and known sites #5127 and Use Spark files to distribute the reference for Spark Walker classes #5221 (since they have not yet been merged) to make passing the reference in Spark more efficient. These changes are needed for the fix below to work.
4. The actual fix is in 3326b90.
5. There's also a new test for ExampleLocusWalkerSpark in addition to the one for CollectAllelicCountsSpark.

The problem was that emitEmptyLoci() in LocusWalkerSpark was not working properly. Any intervals that didn't overlap with reads were being dropped, which meant that those loci were not being passed to the LocusWalkerSpark subclass.

[LeeTL1220]

@tomwhite I have reviewed the CollectAllelicCountsSpark piece of this PR. The changes look reasonable. Thanks for rebasing, as well.

[LeeTL1220]

@tomwhite The other changes should probably be reviewed by someone else on the engine team.

[droazen]

This should not be merged until #5127 and #5221 have gone in first -- treating as blocked until then

[tomwhite]

@LeeTL1220 thanks for the review! I'll work with the engine team to get the other parts reviewed.

[jamesemery]

@tomwhite Could this branch be disentangled from ll_CollectAllelicCountsSpark since the actual fix is in 3326b90 we should probably just pull that out into its own branch and push that through with ll_CollectAllelicCountsSpark which doesn't seem directly related.

[tomwhite]

@jamesemery, yes, I've opened #5248 with the isolated fix (and unit test). I'll leave this open for the CollectAllelicCountsSpark changes, which I'll update when the other changes it depends on are merged.

[tomwhite]

Now that #5127 and #5221 have gone in I've rebased this. Note that it will fail until the fix (which I moved to #5248 on @jamesemery's suggestion) has gone in.