Refactoring RDP method

[ArtPoon]

• carrying out in a separate branch
• percent identity is not being calculated efficiently, redundant calls to a function
• missing some doc strings
• identify deprecated functions / variables

[ArtPoon]

Okay I am done refactoring rdp.py in iss100 branch. Need to write more unit tests. Also could use some comparison data against Windows version @GopiGugan

[ArtPoon]

Some things that I was concerned with during refactoring.

• in calculating the quantity $M$ for the following:
  $$G\frac{L}{N}\sum_{m=M}{N}{N\choose m}p^N(1-p)^{N-m}$$
  (from RDP manual)
  the script was originally using the following:
  

  OpenRDP/openrdp/rdp.py

  Lines 152 to 158 in 3b2181b

  	# m is the proportion of nts in common between either A or B and C in the recombinant region
  	nts_in_a = triplet.sequences[0][coord[0]: coord[1]]
  	nts_in_c = triplet.sequences[2][coord[0]: coord[1]]
  	m = 0
  	for i in range(n):
  	if nts_in_a[i] == nts_in_c[i]:
  	m += 1

This is comparing only sequence A to C, but the inline comment notes that a comparison should also be made between B and C. I have changed this code snippet to the following:

                M = 0  # number of nts in common between (A or B) and C in this region
                p = 0  # proportion of nts in common between (A or B) and C in the entire sequence
                L = len(seq_a)  # length of the entire sequence
                for i in range(N):
                    if seq_a[i]==seq_c[i] or seq_b[i]==seq_c[i]:
                        p += 1
                        if i >= left and i < right:
                            M += 1  # within putative recombinant region
                p /= L

• The RDP Manual notes that sequences A and B are more closely related than they are to C in an UPGMA dendrogram, but we not seem to be running UPGMA.
• There are already functions for working with the binomial distribution in numpy and scipy, it does not make sense to re-invent the wheel so I have replaced this part of the code