Add --n_reads option for specifying a subset of reads to interrogate

[jbeaulaurier]

Sometimes a large set of alignments contains more reads than is necessary to get sufficient insight into the methylation patterns in the data. In this case, it is unnecessary to query all the reads in the alignment and a subsampling of reads would be enough. This would allow users to strike a balance between the amount of data that they need to query and the length of time they are willing to wait for the analysis to complete.