[script.train] Add option to drop incomplete last batch;  Improve documentation of options to explain batch processing

[checkpointer] Make it flexible to directory renames; Fix saving of trainlog in case we are interrupted

[docs,scripts] Separate evaluation script; Improve trainer engine code; Reset network model names to match resource names

[configs] Documented all configuration files; Added script to list/describe/copy configuration files; Re-structured user guide

Minor evaluation fixes

See merge request bob/bob.ip.binseg!11

[script/binseg] In evalpred, allow user to specify the prediction map extension, but defaults to .png

Cast vector types to avoid issues with pytorch > 1.0

See merge request bob/bob.ip.binseg!10