Consolidate old/new clique density selection

[dehann]

Make sure this is using the proper densities/potentials/likelihoods/factors stored in csmc.cliqSubFg -- i.e. make sure we are not using the old potentials list:

• packFromLocalPartials!
• packFromIncomingDensities!
• packFromLocalPotentials!

This came up as part of the LikelihoodMessage.hasPriors discussion (#914)

cc @Affie

[dehann]

HI @Affie , all right so this issue has been resolved with #923 (and few preceding PRs too). I was looking at treeinit a bit and this is what's currently happening on Hex. Note that the "parent starting down solve too early bug" (#924) is also still present.

Here is the new result with .useMsgLikelihooods=false

Here is the result for tree init with .useMsgLikelihoods=true and remember that this now includes the msg.hasPriors story from Friday included as well:

So there are probably still two or three bugs rolling around in there somewhere. I'm going to look at resolving #754 and #924

[dehann]

also note that I deprecated the cliqGibbs function and a few of it's utilities. Think they were just making life difficult. Also note that the internal clique solve order of variables still needs to be updated away from the old btnd.potentials plumbing (#925)