~300k vbank balance updates mostly wasted

[dckc]

What is the Problem Being Solved?

In #7473 / #5896 , we eliminated all vpurse balance updates except module accounts, since the provision pool subscribes to an ERTP balance notifier for its module account.

Unfortunately, there are still a lot of useless module account updates. For example,
@michaelfig notes that every cosmos transaction fee causes a balance update of the agoric1...h3nucl fee module account (aka reserve). I'm not aware of any JS code that's listening for those.

@warner did some analysis of mainnet swingset work up to the agoric-upgrade-12 halt and found 311k VBANK_BALANCE_UPDATE events, including:

address	count
agoric1ae0lmtzlgrcnla9xjkpaarq5d5dfez63h3nucl	298,770

Description of the Design

?

Security Considerations

could become an availability risk at some point

Scaling Considerations

yes :)

Test Plan

?

Upgrade Considerations

?

[dckc]

I asked @warner about impact...

Relatively low, mean 17ms / 127kc (computrons) each. So we do 500 before we fill a block, and could do 100 before making the compute time be significant. It adds to the growth rate of DBs, so I'd like to reduce it, but the actual CPU cost is not huge at the current rate.