Bulk Search API improvements

[fbanados]

The Bulk Search API is core to the sound features of our morphodict instances. However, it seems to be taking a considerable toll on the server, so at least the following should be attempted:

• Evaluate performance optimizations that could improve the behaviour of bulk search (cache, indexing, etc.) to reduce load on server
• If that doesn't work, consider other options.

[fbanados]

Current reliance on iregex for inexact searches of vowel lengths is a bottleneck.

A simple-ish solution would be to perform a kind of iregex search once when a sentence is imported into morphodict and then use exact search instead, which would considerably improve performance.

Secondarily, sorting should be done at the database level, with an index.

[fbanados]

Problem relates to solution for #462

[fbanados]

Switched to exact search. Next step is to avoid the sorting requirement for responses. Then indexes.