Remove masked sites from lineage imputation requirements

[hyanwong]

We have masked out a number of sites from the sc2ts ARGs, but I suspect they haven't been removed from the definitions used to impute the pangolin status of internal nodes. If so, I think this might be screwing up the pangolin imputation. I guess the easiest thing is to remove them from each PANGO lineage definition after doing lineages.read_in_mutations(). Would that be enough, @a-ignatieva, and would it mess badly with imputation?

Here are the ones I can find:

>>> np.array(list(sorted(read_in_mutations("../data/consensus_mutations.json").all_positions - set(ts.sites_position))))

array([   21,    38,    44,    61,    66,   100,   101,   105,   106,
         136,   147,   149,   157,   168,   174,   180,   181,   187,
         190,   193,   199,   203,   204,   207,   209,   210,   219,
         221,   222,   224,   225,   228,   233,   239,   241,   242,
         246,   248,   253,   257,   823,  1191,  1684,  1820,  1912,
        3096,  9430, 10138, 10323, 10369, 11083, 11750, 12439, 14724,
       15240, 15738, 16887, 17676, 19524, 19983, 21077, 21137, 21304,
       21575, 21614, 21846, 21987, 25521, 25904, 25916, 26681, 26801,
       26895, 27384, 27389, 27476, 27638, 27925, 28077, 28087, 28095,
       28253, 28254, 28378, 28603, 28657, 29095, 29543, 29614, 29675,
       29686, 29687, 29688, 29690, 29691, 29692, 29700, 29702, 29705,
       29706, 29708, 29710, 29711, 29718, 29719, 29724, 29726, 29728,
       29730, 29734, 29736, 29737, 29738, 29739, 29741, 29742, 29743,
       29744, 29745, 29747, 29748, 29750, 29751, 29754, 29755, 29762,
       29764, 29766, 29769, 29771, 29774, 29777, 29779, 29781, 29782,
       29784, 29791, 29810, 29820, 29834, 29835])

[hyanwong]

The easiest way to get a list of excluded sites, I assume, is to do what I did above and look to see if they are missing from the inferred ts.