Address issues with heuristic placement

[pierrebarbera]

For "weird" sequences that don't fit in the tree, or may not align correctly, or are just plain wrong, we have observed a tendency of them being placed on longer branches of a reference tree, sometimes with a high to very high LWR. While some of these may be caught by filtering based on pendant length of the queries, the real problem lies with with the heuristic preplacement phase which is the likely culprit. Specifically, during this phase queries are inserted using a default pendant length of 0.9, which for some cases may simply be too long.

This also touches on identification of "novel" lineages in the query data, which is usually a goal of placement analyses. However the primary goal is to re-establish LWR as the primary criterion for placement confidence.