Use timeout for heavy rpc handler

[yihuang]

Proposal: Use timeout for heavy RPC handler

Current behavior:

Some rpc endpoints are heavy:

• debug_traceTransaction
• eth_call
• eth_estimateGas
• eth_getLogs
• etc.

Right now, only eth_call is run under context.WithTimeout, better to implement a timeout mechanic for all heavy APIs, especially the trace transaction, which could cause a delay in node syncing.

Desired behavior: apply a timeout to all heavy endpoints.

Use case: more stable public rpc service.

Requests may be closed if we're not actively planning to work on them.

[yihuang]

var DefaultHTTPTimeouts = HTTPTimeouts{
	ReadTimeout:  30 * time.Second,
	WriteTimeout: 30 * time.Second,
	IdleTimeout:  120 * time.Second,
}

These are the default timeout values for the HTTP server in go-ethereum.
I guess we should make it configurable for public rpc servers.