chore: replace os.path.join with tf.io.gfile.join

[adriangb]

Partially addresses #15300

I only replaced some uses of os.path.join.
I didn't want to replace it where not needed (it is only needed in things that interact ram:// or gcs:// filesystems). But I probably replied it in some places where it's not needed and neglected to replace it in places where it is needed. The codebase is very large and I'm unfamiliar with a lot of it; I'd appreciate some guidance from someone with more global knowledge of the codebase.

[google-cla]

We found a Contributor License Agreement for you (the sender of this pull request), but were unable to find agreements for all the commit author(s) or Co-authors. If you authored these, maybe you used a different email address in the git commits than was used to sign the CLA (login here to double check)? If these were authored by someone else, then they will need to sign a CLA as well, and confirm that they're okay with these being contributed to Google.
In order to pass this check, please resolve this problem and then comment @googlebot I fixed it.. If the bot doesn't comment, it means it doesn't think anything has changed.

ℹ️ Googlers: Go here for more info.

[adriangb]

@gbaned can you kick off CI please?

I'll give this another pass in the next couple days, I can already see plenty of places I don't think replacement was needed

[adriangb]

I went through this by eye. The main change I made was to revert (i.e. make no changes) to callbacks.py. It seemed like most of the path joins there were to write out logs, and it didn't seem like the ram:// or gs:// filesystems would be valid in that context. Please correct me if this is wrong.

I think all of the replacements left now directly pertain to SavedModel, which is the context where ram:// and gs:// are certainly valid and this change makes sense.

[fchollet]

LGTM, thank you