Support custom tags to neurodata types

[rly]

Instead of going hierarchical & folder-y, another way I was thinking of was to tag these objects, which would be more powerful because certain time series or tables might have information about both stimulus and behavior (though rarely). Plus, such tags could then be used to query for all objects spanning across acquisition, processing and analysis, (even lab-meta-data if needed). However, except for the trials table, this does not seem feasible.

from NeurodataWithoutBorders/helpdesk#37 (reply in thread) @tuanpham96

I like the idea of tagging objects (neurodata types) with custom strings and agree - it would be more powerful than the hierarchical/folder-structure approach.

To implement this for all neurodata types, I suggest adding a 1-dimensional string attribute called "object_tags" to NWBContainer and NWBData which are the base types for all NWB groups and datasets. And we could add a convenience function in PyNWB that returns all objects with a given object tag.