Filter mapping file issue

[wdwvt1]

This PR does 3 things:

1. Addresses issue #2061: It changes the way qiime.filter.filter_mapping_file behaves to match the documentation.
2. Changes the result --output_mapping_fp in filter_samples_from_otu_table.py: now writes a subsetted mapping file that retains all the metadata columns of the original passed mapping file regardless of their uniqueness status.
3. Adds a script usage test for the script filter_samples_from_otu_table.py for the --output_mapping_fp

I believe change 2 is good because: we are not exposing the control of whether or not unique metadata columns are written in the script (we don't mention it in the documentation either) and so it is very startling to see that data disappear when you use the --output_mapping_fp option. In addition, its easy to remove columns in e.g. excel, whereas its annoying to have to add columns back in if you have output a reduced mapping file. Also, its useful to retain non-unique and totally unique columns with a reduced mapping file if you are e.g. trying to work with a new dataset and see what the values for a particular set of samples are.

[ghost]

Build results will soon be (or already are) available at: http://ci.qiime.org/job/qiime-github-pr/1663/

[gregcaporaso]

retest this please

[ghost]

Build results will soon be (or already are) available at: http://ci.qiime.org/job/qiime-github-pr/1719/