TS-GRG: Consider support for infinite-sites modeled mutations

[dcdehaas]

msprime supports discrete_genome=False for sim_mutations(), which causes the positions of the mutations to be in floating point. GRG initially supported the Mutation position to be double, which would have allowed this to work when converting a tree-seq to a GRG. At some point we changed the position to be an integer.

Just an interesting use-case to consider. One possibility is to map the resulting mutations from TS onto the integer number line. Another is to consider re-adding (optional) floating point support in GRG positions (probably by just adding a special getter that would interpret the value as float)