Energy metering for persistent memory usage

[coolreader18]

Description of Changes

As on the tin. I think this is the idea, right? Or am I missing another part of the database that stores user data in-memory.

Expected complexity level and risk

2

[Centril]

Also, could you please integrate this in spacetimedb_table? I'd like to avoid mem-arch getting further away from being merged.

based on mem-arch

[bfops]

@coolreader18 I just want to confirm for PR shepherding purposes - this change is backwards-compatible right?

[coolreader18]

@bfops yes, it is.

[coolreader18]

Oh, from the user side at least - cloud will need an update (but I have it ready already)

[coolreader18]

The hashbrown update is because 0.15 added an allocation_size() method, which is obviously very helpful. It also changes the default hasher, though, and I didn't want this to affect anything else, so I set our default hasher back to ahash, and all the HashCollectionExt imports are because we can no longer use HashMap::new().

[joshua-spacetime]

Thanks @coolreader18! Just a few typos, but otherwise looks good.

You've accounted for nearly all mem usage :)

I think it would have been sufficient to just track the pages, indexes, blob store, and pointer map. That probably could have reduced a lot of the boilerplate. But this is very nice nonetheless.