Add implementation of optimal hash size calculation to khmer.khmer_args

[ctb]

@qingpeng, can you take a first stab at adding a function into khmer to calculate the optimal H and Z parameters for a CountMin Sketch at a given fp rate for a given N? Don't worry about argparse inclusion or anything -- just a function with some unit tests.

[qingpeng]

OK.

[ctb]

Actually, reading through the section in the paper, I think we should have two functions: one that calculates the correct Z and H given M (-N and -x given desired memory usage) and one that calculates the Z and H given f and N. Then when the HLL comes on line (hah!) we can make use of the latter. /cc @luizirber

[mr-c]

@qingpeng & @luizirber Any update on this?

[ctb]

Scenarios --

• not useful for digital normalization, where we want to know how many k-mers are left AFTER diginorm;
• useful for partitioning

Anything else?

[mr-c]

Still useful for diginorm as it can provide a lower ceiling for memory usage if number of k-mers is previously known.

[proteasome]

@mr-c i'm interested in working on this

[ctb]

Added in #1106.

[ctb]

Fixed by #1126.