Publish amm pool list subscription to chain storage

[michaelfig]

refs: #4398
refs: #5386

Description

Created new makeStoredSubscription to produce a StoredSubscription<T> (a subscription with Merkle tree query information). Passed a storageNode and marshaller from bootstrap to the AMM contract for a new StoredSubscription<{ XYK: Brand[] }>.

Drive by:

• reenabled CI build cache
• allow synchronous throws in iterationObservers to terminate the iteration
• fix makeScaledPriceAuthority to query its underlying authority just-in-time

Security Considerations

Documentation Considerations

Testing Considerations

Here's a demo:

snow:cosmic-swingset michael$ agd query swingset keys ''
keys:
- activityhash
- published
pagination: null
snow:cosmic-swingset michael$ agd query swingset keys published
keys:
- amm
pagination: null
snow:cosmic-swingset michael$ agd query swingset keys published.amm
keys:
- metrics
pagination: null
snow:cosmic-swingset michael$ agd query swingset keys published.amm.metrics
keys: []
pagination: null
snow:cosmic-swingset michael$ agd query swingset storage published.amm.metrics
value: '{"body":"{\"XYK\":[{\"@qclass\":\"slot\",\"iface\":\"Alleged: IbcATOM brand\",\"index\":0}]}","slots":["board0371"]}'
snow:cosmic-swingset michael$

[dckc]

fun stuff!

just a few tweaks to consider

[dckc]

maybe change the "TODO: magic string" comment to a justification: 'metrics' is derived from getMetrics which is already part of the public API. I suppose it's still feasible to make a manifest constant.

[michaelfig]

I left this TODO intact, for a later date.

[mhofman]

This was needed for the deployment test job which couldn't use the build cache because it runs on a different linux environment, so the native modules would not match. The default value of true should have made it no impact on any other job. I need to understand why you claim it affected other jobs.

[michaelfig]

I'll create a new PR to distinguish the cache key without an explicit toggle.

Objectionable change was removed from this PR.

[gibson042]

@michaelfig question: how does the code calling observeIteration differentiate between a failure in the iterator and an error thrown from its own listener?

[michaelfig]

Not currently possible. The right behaviour would be: if you install a fail handler, it will only be called if the iterator publishes a failure. Listener throws should only be reflected by rejecting the outermost promise, and not call the fail handler.