feature request: Max/min line of sight separation for summing pairs in new rperp metric about treecorr

[rmjarvis]

Done on master. Rlens and Rperp metrics are allowed to give min_rpar, max_rpar parameters which limit the minimum and maximum value of r_source - r_lens, where "source" is the second catalog of the two, and "lens" is the first catalog.

This is a signed quantity, so to specify that you only want background galaxies, you can use min_rpar = 0. To specify that you only want local galaxies (for II or halo studies) you can use e.g. min_rpar = -50 and max_rpar = 50.

One caveat is that I don't use the min/max rpar values for splitting, just for pruning pairs of cells. This means that if you want a fairly exact rpar cutoff, you might need to use a lower than normal bin_slop to make sure that the regular splitting process gets down to small enough cells that the error in rpar is small. OTOH, I suspect most people won't need the rpar cutoff to be extremely precise, so I think this is fine.

The usual recommendation about bin_slop applies: Check if using a bin_slop half as large makes any significant difference to the statistic you are making, and if not, then you're fine. If it makes a difference, you probably need to go smaller.