Remove nbytes_in_memory

[mrocklin]

This doesn't appear to be necessary for any scheduling decisions today, and is causing issues. Let's remove it.

[mrocklin]

git blame puts the blame on me on this one. This was added as part of the original TaskGroup/TaskPrefix PR, which did carry the sin of speculative work. My apologies to all.

[gjoseph92]

I was thinking of a similar solution 😄

What about nbytes_total? I'm not sure what that's supposed to mean, but it might have similar problems.

I could see nbytes_in_memory become useful in the TaskGroup visualization, but if we actually want it we can add it back for that.

[mrocklin]

I recommend that we wait on merging this PR until after the release today, just in case someone was depending on this internal state.

[mrocklin]

In the meantime, cherry-picking this PR might be a good idea :)

[jakirkham]

Also happy to see this go away. I don't think it was used anywhere

[jakirkham]

Out of curiosity what issues was it causing?

[jrbourbeau]

This happened to pop up and cause issues in a couple of places over the past few days. #4925 (comment) is one of those places. #4927 has a nice writeup from @gjoseph92

[gjoseph92]

#4927, #4927 (comment), #4925 (comment). Oddly, after a long period of dormancy, a few of us all ran into this from different angles at the same time.

[jakirkham]

Ok thanks all 😄