Remove query cancellation from StateManagers (was: comment out all instances of manual query cancellation

[bcolloran]

experiment with removing query cancellation

Here's what I see in a couple conditions

(1) state as of changes in this PR:

• all manually query cancellation removed
• observeContextColumnWidth reactive statement commented out

Results:

• everything above the fold loads in about 15 seconds,
• loading keeps going for stuff below the fold
• no queries cancelled

https://github.com/rilldata/rill/assets/2380975/5ed313ac-b47a-4fed-84d0-6d3f0fd163d6

(2) Main before this PR, but without observeContextColumnWidth

• manual query cancellation in place
• observeContextColumnWidth reactive statement commented out

Results:

• everything above the fold loads in about 15 seconds,
• loading keeps going for stuff below the fold
• no queries cancelled https://github.com/rilldata/rill/assets/2380975/194cfdfc-769a-4fdb-9acb-ad7b5d115205

[bcolloran]

@AdityaHegde and @djbarnwal would love to get your takes on this also before we go forward with any changes.

@ericpgreen2 and I have been looking at query cancellation in the context of https://github.com/rilldata/rill-private-issues/issues/112, and think that we might be doing extra manual cancellations in places where tanstack should be able to do that automatically. I think that as a team, we need to audit all of those manual cancellation and understand why they are needed, and if they can be removed.

@AdityaHegde, I've assigned this to you because I think you probably know the most about the intersection between runtime and client here. I think "done" on this would be code comments for any places we want to retain manual query cancellation. I don't mind making the changes in this PR if you just want to put those comments as review comments.

[djbarnwal]

I haven't personally worked a lot with query cancellation and relied on tanstack for it. I am up for removing manual cancellation if they don't have any performance implications.

[bcolloran]

@ericpgreen2 cleaning this up to get it over the line -- removed changes in ModelBody.svelte and resource-invalidations.ts, per you and @AdityaHegde's suggestions