Enhance assembly profiles

[linsalrob]

At the moment we have a simple assembly approach:

1. Assemble each sample separately
2. Map reads
3. Identify unassembled reads
4. Repeat

There is another approach we might consider:

1. Assemble x% of all reads (e.g. 0.1% or 1% depending on how big the data sets are..could this be dynamic?)
2. Map reads and find unassembled reads
3. Double the % of reads assembled and go to step 1
4. Keep iterating this until we have assembled everything

Each approach has its pros and cons, of course, but the 2nd approach might remove all the redundant stuff in all the libraries quite quickly.