cqlsh could adjust client request timeout for USING TIMEOUT queries

[avelanarius]

This is a small suggestion for a user experience improvement. In Scylla 4.4.0, USING TIMEOUT syntax was introduced. However, if you want to execute query with per-query timeout in cqlsh, you must remember to change --request-timeout option:

$ cqlsh 127.0.0.2
cqlsh> SELECT COUNT(*) FROM ks.zrodlo USING TIMEOUT 20m;
OperationTimedOut: errors={'127.0.0.2': 'Client request timeout. See Session.execute[_async](timeout)'}, last_host=127.0.0.2

$ cqlsh 127.0.0.2 --request-timeout 1200
cqlsh> SELECT COUNT(*) FROM ks.zrodlo USING TIMEOUT 20m;

 count
---------
 9995165

(1 rows)

For better user experience, cqlsh could:

1. Give a better error message. Currently it prints:

OperationTimedOut: errors={'127.0.0.2': 'Client request timeout. See Session.execute[_async](timeout)'}, last_host=127.0.0.2

and there is no suggestion to use --request-timeout flag.

2. Automatically change client request timeout (or maybe ask the user if they want to change it). I am aware that to do this it could require parsing the query on the cqlsh side (so possibly not worth it).

/cc @psarna (only a suggestion, not that this is important)

[psarna]

@avelanarius cqlsh already parses statements: https://github.com/apache/cassandra/blob/56ea39ec704a94b5d23cbe530548745ab2420cee/bin/cqlsh.py#L970 , and even though USING TIMEOUT is ScyllaDB-specific, USING isn't on its own (https://github.com/apache/cassandra/blob/f28dd90feb215db85ac2e510c5657a49edd46e12/pylib/cqlshlib/cqlhandling.py#L32), so there's a fair chance such validation can be done rather easily.