allow loading from treeSeq even when treeSeq isn't enabled?

[petrelharp]

Currently, trying to do sim.readFromPopulationFile() raises an error if you haven't done initializeTreeSeq(). Sometimes people want to run simulations without tree sequence recording, for speed, and so do things like

c.     Wrote a script to take the treesequence at time 0 and write a file in slim output

to move from a treeseq simulation to one without. However, I don't know of any reason that you shouldn't be able to load in a simulation from a tree sequence and yet not have tree sequence recording turned on? This would make it possible, for instance, to intiialize a giant simulation from an msprime simulation and run it without recording the tree sequence (for speed).

[bhaller]

Hmm. If they want to do that, they can just dump the msprime output in MS or VCF format and load that in, no? I guess you could make the case that the .trees file will be smaller, faster to load in, and express the pattern of mutations more concisely, so for a giant simulation perhaps it would work where MS/VCF wouldn't. Hmm. Well, I guess a .trees is "just another file format", in a sense, and there's no particular reason why users shouldn't use it to ferry data around even if they don't intend to use tree-seq recording in SLiM. And I guess we can just (a) pretend that tree-seq recording is on, (b) load the .trees as usual, and then (c) throw all the tree-seq data structures away. For some reason this makes me feel nervous, but I don't see a problem with the idea offhand...

[bhaller]

OK, this seems to work now. I haven't tested it extensively, but I don't really see why there would be any weird edge cases; we just load and then throw away the tree-seq data structures. Let me know if you see any problems.