mock-node: simplify and remove client setup code

[marcelo-gonzalez]

Before this change, the mock-node tool starts one mock node, and one normal neard node with nearcore::start_with_config(), and then waits for the node to sync. It also modifies and sets up a home dir for this client if --client-height is given on the command line.

However, nobody was really using this --client-height feature much. Since we can get good test/benchmark results from just using a regular neard home directory taken from mainnet or localnet testing, we don't really need any code that manually sets up the database state.

Then if we get rid of that, the question arises: why does this code even start a node with nearcore::start_with_config()? There's not a great reason to do it in the same binary, and doing so restricts the use of this tool, since we can't use it to test a standard neard binary, but must build a custom mock-node binary that's not very intuitive unless you've looked at the code.

So here we delete all the client set up code and just have the node start a mock node listening on whatever addr is specified in the near config. Then to use it to benchmark the sync performance of a node, you just need to set the right boot nodes argument (and possibly also manually wipe the peer store), and you can use a normal neard binary for tests

[marcelo-gonzalez]

Not really sure who to mark as a reviewer, since not many have been touching or looking at this lately, but @wacban looks like you are the lucky random assignee :D

[wacban]

Let me just stamp it for now assuming it's a prototype. Once we've agreed that this is something we want to include into our benchmarking / debugging tool suite let's do a proper review of the whole thing.