skip document during indexing if error occurs

[tonysun83]

When an error is returned by search3-java, we handle it by throwing a
generic bad_request exception. This is the error handling path for both
search and indexing requests. During indexing however, the exception is
not caught and it leads to a search_worker_manager process crash. The
process is restarted by the search3_sup supervisor and the indexing
begins again. This leads to an endless loop of trying to index the same
document if the documents causes some sort of error. We modify this
behavior by catching the thrown exception and ignoring the document.

[rnewson]

Sidenote: verify_same_session is more idiomatically written as

% Session Verification
verify_same_session(Session, Session) when is_binary(Session) ->
    Session;
verify_same_session(_, _) ->
    throw(session_mismatch).

[rnewson]

this doesn't look like the right result. handle_response returns {VSession, Response} above, for example.

[tonysun83]

@rnewson handle_response was a generic handler and Response was need for other rpc requests. However in the case of updates, Response isn't needed, so I removed it.

[rnewson]

+1