inference ll != test ll about sts

[matsen]

OK, well, the problem only appears to be with the delta bl density.

stoke re/sts ‹59-infer-ll› » ./_build/sts -p 197 --bl-dens delta data/bppsim/cherry/seq.fasta | sort -n -r | head -n 1
Iter 1 max ll -312.286 ESS: 197
-312.286        (s1:0.901385,s4:0.901385);
stoke re/sts ‹59-infer-ll› » ./_build/sts -p 198 --bl-dens delta data/bppsim/cherry/seq.fasta | sort -n -r | head -n 1
Iter 1 max ll -38.7707 ESS: 198
-38.7707        (s4:1.09621,s1:1.09621);

versus

stoke re/sts ‹59-infer-ll› » ./_build/sts -p 197 data/bppsim/cherry/seq.fasta | sort -n -r | head -n 1
Iter 1 max ll -255.234 ESS: 4.18316
-255.234        (s1:0.0651214,s4:0.174163);
stoke re/sts ‹59-infer-ll› » ./_build/sts -p 198 data/bppsim/cherry/seq.fasta | sort -n -r | head -n 1
Iter 1 max ll -255.238 ESS: 4.18316
-255.238        (s1:0.0641175,s4:0.173159);

But what is strange is that the problem is really just when the
online_calculator has to grow. Hm...

[matsen]

Scratch that. Not just a problem with delta.

stoke re/sts ‹59-infer-ll› » ./_build/sts -p 198 data/bppsim/cherry/seq.fasta | sort -n -r | head -n 1
Iter 1 max ll -255.238 ESS: 4.18316
-255.238        (s1:0.0641175,s4:0.173159);
stoke re/sts ‹59-infer-ll*› » ./_build/sts -p 199 data/bppsim/cherry/seq.fasta | sort -n -r | head -n 1
Iter 1 max ll -38.8162 ESS: 1
-38.8162        (s4:1.14416,s1:6.92983);

[matsen]

@cmccoy and @koadman -- I'm not going to pursue this bug any further for a bit. The problem only happens when the online_calculator has to grow. I'm not so familiar with what problems might lurk there, but may take another look this PM.

Will be great to have this one squashed!

BTW, it goes all the way back to 45468e9