v0.50.0

Added

• The genome rates can vary among lineages.
  duplication=PerCopy(0.1).varying_among('lineages', LogNormal(0.0, 0.4)) draws one multiplier for
  each branch of the species tree, before the run starts. That multiplier scales the rate of every
  family on the branch. varying_among('lineages', Drift(...)) is the inherited form, where a
  daughter branch starts from its parent's multiplier and takes one step from it. Both forms work on
  duplication, transfer, loss and origination. On transfer the multiplier belongs to the donor, and
  transfer_to still chooses the recipient. Several rates can share one multiplier by sharing one
  Random('lineages', ...), so a fast branch is fast at all of them. The multipliers are written in
  lineage_multipliers.tsv, one row per branch and one column per rate, and .lineage_multipliers
  holds them in Python. A parallel= run draws them before its workers start, so a run gives the
  same multipliers for any worker count; stream_to= writes the file too. (#447)
• The ordered resolution takes the same draw, on every rate it has: duplication, transfer, loss and
  origination, the three rearrangements and the four chromosome events. A branch's multiplier scales
  whatever the event acts on. A family's multiplier applies to the genes a segment covers, so the
  per-family draw is on the per-gene events only. The ordered lineage_multipliers.tsv has a column
  for each of those eleven rates. A per-lineage draw and a per-family draw work together in one
  ordered run; a per-family draw and a driver still do not. The nucleotide resolution does not take
  the draw, and neither does an extent, at any resolution. (#447)