Render dynamic UI with prodash

[stuhood]

Switches to rendering the --dynamic-ui with prodash. Although prodash has fewer users than indicatif, the documentation and internals are very solid, and the project maintainer is highly responsive. prodash's conceptual model is also a much closer fit for our workunits: its core data model is a tree, which naturally allows for rendering workunit hierarchy, and dynamically adding and removing tasks.

This change would resolve #13367 by adjusting the number of swimlanes dynamically. But it would also pave a clear path to #11965 to both render task progress via gauges (to show e.g. download progress, or the expected runtime of processes), and potentially by rendering either a child workunit, or the last line of output for long-running tasks as a child of a task.

[stuhood]

This is a draft both because there are a few more issues to nail down (Byron/prodash#9 and Byron/prodash#10), and also because I think that we'll want to see whether we can actually implement some of #11965 before landing this. There are likely also styling changes to be made (light blue is odd, and folks might miss the spinner!).

[thejcannon]

Hey @stuhood what would it take to get this in Pants under some experimental flag?

I'm thinking maybe [GLOBAL].dynamic_ui is now coerced to a string with the following values:

• false - no dynamic (backwards-compatible)
• true - alias for spinner (backwards-compatible)
• spinner - current spinner impl
• experimental_prodash - this PR

[stuhood]

Hey @stuhood what would it take to get this in Pants under some experimental flag?

Landing this as experimental would require Byron/prodash#9 landing at a minimum (although for an experimental implementation, I suppose that we could land while using the fork).

The largest question is really: does this seem like an improvement for you? Have you been able to try it out and see what you think of the experience of using it? Or will getting that experience require landing it first?

[thejcannon]

Last time I used it I remember it being much improved. I can try getting some recordings to compare tomorrow.

[stuhood]

Last time I used it I remember it being much improved. I can try getting some recordings to compare tomorrow.

Thanks: that would be helpful. I've rebased it.

[thejcannon]

(Sharing videos in Slack)
The difference is stark when < # of available lines processes are executing. Which with 64 cores is often :)

[thejcannon]

@stuhood closing in favor of #14221