the --trace flag is not respected when querying eth JSONRPC -> Cosmos RPC

I was tracing down a bug where SDK code was panicking. when this occurs within the context of cosmos SDK grpc/rpc handlers, i was able to get the full stack trace by starting the node with `--trace`. However, when calling an eth JSONRPC method that calls a Cosmos SDK query that invokes a panic, you do not get the trace.

we should see if we can figure out a way to get panic traces all the way through the JSONRPC layer.