Make blockNumber rpc more efficient with atomic uint64

[rlan35]

Previous _blockNumber rpc is retrieving the latest block, copying the header and returning the number from the header, which is very unnecessary and expensive.

Adding a uint64 atomic value to store the latest block number and directly turning it.

Also increase receiptsCacheLimit to 256 and reduce rpcGetBlocksLimit to 64

Change tested locally.

[rlan35]

How much resources we are saving for blockNumber requests? Is there a benchmark showing the improvement?

No benchmark yet. @AlexiaChen is working on a loadtest and benchmark tool.

[ONECasey]

I really like this PR. I will be resubmitting it.