Optimize transcriptome simulation

[kmnip]

• consolidated code for progress updates for number of read simulated in metagenome/transcriptome/genome modes to a new method check_print_progress
• reduced frequency of writing updates to stdout, e.g. once every 10,000 reads simulated instead of 100
• optmized code for transcriptome simulation

Simulated 500,000 reads using 4 threads on the same machine for 3 trials:

trial	before	optimized
1	1h:00m:13s	41m:42s 16m:38s
2	1h:12m:52s	50m:47s 16m:40s
3	1h:03m:18s	44m:53s 16m:42s

About ~75% reduction in wallclock runtime.

[kmnip]

I have further optimized the code in afe3cbe.

A new sample is only drawn from the KDE if the existing sample was previously used to simulate a read for the candidate transcript.

I have repeated my benchmarking trials on the same machine and updated the table above. The reduction in runtime is now ~75% (instead of ~30%)!

[kmnip]

Added a potential fix for #156

[SaberHQ]

I have further optimized the code in afe3cbe.

A new sample is only drawn from the KDE if the existing sample was previously used to simulate a read for the candidate transcript.

I have repeated my benchmarking trials on the same machine and updated the table above. The reduction in runtime is now ~75% (instead of ~30%)!

This is really a neat and clever optimization @kmnip. Thanks for this. I also benchmarked it by simulating 1 million and 25 million reads using 38 threads on a same machine and the results demonstrates a reduction in wallclock runtime.

Number of reads	before	optimized
1 million	15m:56s	10m:29s
25 million	29h:46m:26s	5h:34m:16s

[SaberHQ]

This pull request potentially fixes #159 and #131