[Enhancement] Add error/noise processes to the simulation.

[reedacartwright]

As this project moves forward, it might be helpful to have some standard ways to modify simulation output to add noise and error to the data. Some of these would require changing the simulation process itself. Others could be implemented as post-simulation modifications. E.g.

• A rate of technical errors that cause variants to be mistyped or missing
• A parameter to inject noise into the coalescent process, such as making branch lengths more variable than they would normally be.
• Or similarly, make mutation rates vary along a chromosome (or time).

The idea here is to not try to create meaningful models of a biological process, but provide a way to introduce random noise at different parts of the simulation in order to benchmark robustness of analyses.

[gtsambos]

A parameter to inject noise into the coalescent process, such as making branch lengths more variable than they would normally be.

See #299 for a bit of existing discussion about this.

Or similarly, make mutation rates vary along a chromosome (or time).

I think this would be a useful change for msprime as a whole -- from memory I think @carjed was interested in this too??

[jeromekelleher]

Adding support for error would be excellent, and really useful in benchmarking methods. Needs some careful thought about how to do it properly though I think.

I think this is a separate issue to #299 though @gtsambos, as that's about pulling parameters from a known distribution rather than adding noise to the output.