Stable placement order

[sjanssen2]

Hi @antgonza,

the current pplacer binary, shipped with SEPP for fragment insertion, throws segmentation faults on newer Debian systems and was already reported two years ago smirarab/sepp#121

My workaround is to make SEPP switch to a later version of pplacer, but this will write the produces placements with a different field order. Considering the qiita archive, I fear that this would produce new placements which are incompatible with the db content :-/

One SEPP run produces a set of placements and fortunately, this file also reports field order like here:

qp-deblur/support_files/sepp/placements.json

Lines 4603 to 4607 in efd59e3

	"version": 1, "fields": ["edge_num",
	"likelihood",
	"like_weight_ratio",
	"distal_length",
	"pendant_length"]

but this is per run and not per placement.

I therefore came up with this PR, that should ensure that placements are read and stored in a specified field order, i.e. rearranges if necessary.

Unfortunately, the current tests fail - but I think this is not due to this PR but a more general issue, see #67