nd_types and mapping to approach/technique

@bendichter, @yarikoptic

this file (`curl -o sub-P10HMH_ses-20060901_ecephys+image.nwb https://dandiarchive.s3.amazonaws.com/blobs/284/eb3/284eb346-0bc5-42a6-9b33-268e6b0b0bde`)

contains the following nd_types.

```
'nd_types': ['AnnotationSeries',
'Device (19)',
'DynamicTable',
'ElectrodeGroup (18)',
'OpticalSeries',
'Subject',
'TimeIntervals',
'TimeSeries',
'Units']}
```

but several are not in the table here and hence skipped in our extraction: https://docs.google.com/spreadsheets/d/1x6aI5e7uWwyUJf_S1RyYkmOW-OzodLcRserZVq4DL2A/edit#gid=0

in the ideal world i would love an nwb function that takes the list of nd_types in a file and returns a list of structured objects relevant to search. so a better version of this:

https://github.com/dandi/dandi-cli/blob/78279cc83af5cbbfe98f421e7ff602d277ee1c60/dandi/metadata.py#L640
(the ideal version would also include URIs)

where and how do you think is the best place to create and maintain the mapping table. for the moment i'm happy to continue hacking in the same file, but doesn't seem like the right place. also as yarik has suggested it would be good to move the long dict out to a json file.