Incorrect behavior for fetch_nwb in Linearized position tables

[samuelbray32]

Describe the bug

• Both LinearizedPositionV1.fetch_nwb() and IntervalLinearizedPosition.fetch_nwb() return a string of the absolute path of the analysis file rather than an open nwb object
• Causes additional errors within fetch1_dataframe() for these tables

To Reproduce
Both of these return an array of strings rather than open files

from spyglass.linearization.v1 import LinearizedPositionV1
key2 = LinearizedPositionV1().fetch("KEY")[0]
(LinearizedPositionV1() & key2).fetch_nwb()

from spyglass.linearization.v0 import IntervalLinearizedPosition
key2 = IntervalLinearizedPosition().fetch("KEY")[0]
(IntervalLinearizedPosition() & key2).fetch_nwb()

[samuelbray32]

Correction: Is not specific to the Linearized Position pipeline. TrodesPosV1 fails as well. Probably a recent commit issue

[samuelbray32]

Checking prior versions, error was introduced in #898

[samuelbray32]

Got it.

Removing this line in #898 caused the later projection in this file to not capture the relevant object id's

Also, not forcing return as dict here causes the returned structure from fetch_nwb to be different. @edeno is there a reason we need that change?

Edit: wrt second point. forcing return dict as true is necessary for current implementations of fetch1_dataframe

[edeno]

@CBroz1 actually introduced those changes in a different PR. I just copied them over.

[samuelbray32]

👍

I made a pr that gets it working on the master branch again at #913. Once Chris has a chance to check it doesn't affect new features can merge it in