Create tool that reduces file size for sharing and debugging

[rly]

If there is an error with reading or validating an NWB/HDMF file and the file is large (>1 GB), sharing that file is burdensome, and rarely is the error due to the size of the datasets. It would be useful to create a simple tool to reduce the size of large (>1k elements) datasets and repack the new file to make it easier for sharing. Then when a user has a large file that cannot be read or validated, we can ask them to run this script on the file and share the resulting smaller file.

pip install nwbtrim 
nwbtrim myfile.py -o myfile_trim.py

[oruebel]

I don't think the h5repack utility https://support.hdfgroup.org/HDF5/doc/RM/Tools.html#Tools-Repack can quite do this out-of-the-box, but I think it can still be useful here. I think a simple solution would be to:

1. Copy the file
2. Shrink all large datasets (e.g., using h5py.visititems with h5py.resize)
3. Call h5repack