table selection and parallel batch imports

--only lets you run just a subset of a named configuration's tables (comma-separated db.table names), failing fast on unknown names and preserving config order; it applies to both real and --dry-run runs, with unset meaning all tables as before. --num-threads runs a batched table's per-batch INSERTs concurrently across N numbered workers, raising throughput while keeping the default of 1 sequential and bit-for-bit unchanged. The per-table loop stays sequential and the first failing batch cancels the rest, and with more than one thread each progress line is annotated with the worker that pulled that batch.

Batch copy

Copy data in batches and remove restriction of table size to be dropped.

Control Timeouts

Added a top-level connection: block to the YAML with dial_timeout and read_timeout, wired through to the Go ClickHouse client so long-running imports aren't cut off by the 5-minute default ReadTimeout.