Fixes #43

This is hopefully a temporary measure. At the moment we cannot accurately
predict the running time of the extractor, so we have to aggressively
limit the input based on the worst-case expectation.

As a worst-case scenario, we currently use random input sequences, where
length 1000bp yields about 500ms of running time.

In the future we hope to either:

1. Predict the running time and abort if needed.
2. Keep track of the running time and abort if needed.
3. Run the extractor in a task scheduler with a running time limit.

Jeroen F.J. Laros authored Apr 25, 2015 and Vermaat committed May 18, 2015

We can now compare two sequences by supplying their sequence strings,
accession numbers, or uploaded file.

This is a work in progress as there still seem to be some bugs. For example,
some unit tests fail due to incorrect descriptions generated and others fail
due to a crash.

This includes changing a lot of routes and parameter names to be more
consistent. We try to remain backwards compatible as much as possible
by providing redirects from old routes and parameter names.