Apply timeout to persistent connections

[camfilho]

The persistent socket is built with HTTP.persistent(...) alone, so the timeout option only reaches the non-persistent path in execute_request. Since persistent: true is the default, most clients make requests with no client-side timeout:

client = SerpApi::Client.new(api_key: 'x', timeout: 5)
client.socket.default_options.timeout_options # => {}

A request on a half-open connection then waits on the OS TCP retry limit, and a server that accepts but never responds can block the caller indefinitely.

Change: build the socket with HTTP.timeout(timeout).persistent(...). http.rb resets the global timeout counter after each response on a kept-alive connection, so long-lived clients are not cut off after timeout seconds in total.

The README options table describes timeout as applying to non-persistent requests only, while its examples pass timeout: with persistent: true. I've updated the table to match; if the current behavior is intended, happy to close this.

Tested:

• New spec applies the timeout to the persistent socket fails before the change (got: {}) and passes after. It makes no network call.
• rubocop lib spec: no offenses.
• Two live searches on one persistent client with timeout: 30: both succeeded over the same socket, and close returned :clean.

This fix was also part of #22, which was closed by its author along with several unrelated changes. This PR only carries the timeout fix.