getlogs and internal timing polish

[shunjizhan]

Context

we have a bug that returns false empty for the corner cases when:

1. get log immediately on new block, since subql is usually a couple blocks delayed
2. get receipt immediately on new block, since cache is still syncing latest block

1. was temporarily fixed by #828, but it has very poor performance.
2. was not fixed yet

To solve the whole issue completely, we need some bottom level refactor, instead of awkward patches like #828

core problem

We directly use node best block as eth best block, so we can query for this block before subql or cache is ready. In other words, latest block goes faster than latest state.

solution

We can redefine "latest block" as "latest cached block", so we can always get logs/receipts for the latest block.

For logs, we can't wait for subql to catch up, which takes 10+ seconds, instead we can make use of cache to query the blocks missed by subql

Change

• updated blockcache so we can get logs from it, and it keeps track of last cached block info
• updated latest block to return latest cached block
• now we should always be able to get receipts and logs in latest block since:
  • latest block info is guaranteed to be in cache
  • there is a check to make sure no gap between subql and cache

fix #863

Test

• updated some tests to wait for eth rpc to report new block before proceeding
• add tests to make sure we can get logs as well as receipts for the latest block as soon as the block is ready