mariadb-dump/import: parallel partitions

Currently, mariadb-dump -T can dump multiple tables in parallel threads, but every single table is always dumped sequentially. For a big table with lots of partitions it makes sense to dump individual partitions in parallel.

• Dump file naming: after MDEV-37483 file name is an encoded table name, as as in the server, so using #p# partition suffix — same as in the server — sounds like an obvious and safe choice.
• Importing: two options here, mariadb-import can ignore everything after #p# suffix and load files into the same table — the server will figure out what partition the data belongs to. But a more concurrently performant approach would be to load every partition into a separate table and then use ALTER TABLE ... EXCHANGE PARTITION to put fully populated partitions in place.
• SQL syntax to use in mariadb-dump internally:

  SELECT * FROM table_name PARTITION (partition_name)

• what option to use to enable it? — don't. Simply treat all partitions in all tables as separate entities and schedule them all to be dumped in parallel according to the concurrency settings of the thread pool.