Use number_of_read_tuples param when greater than coverage's

[simonepignotti]

In the current implementation, every simulator apart from ArtIllumina has mutually exclusive number_of_read_tuples and coverage parameters.
When the genome is short and coverage is low, this result in no read being simulated (and, currently, in error #53).

A desirable behaviour would be to use the max between number_of_read_tuples and the effect of coverage as final value for the number of reads.
This would guarantee that at least number_of_read_tuples are simulated for each fasta. By default, number_of_read_tuples should be one.