Partition temporary (and resulting) vector index tables before uploading

[kunga]

Because we use (parentClusterId, ...) keys in build and resulting tables we may evenly them create with necessary number of partitions before upload to speed up build index process

It seems fine to make 1 shard is twice bigger than expected evenly distributed rows among clusters

[kunga]

Seems already done for temp build tables here

ydb/ydb/core/tx/schemeshard/schemeshard_build_index__progress.cpp

Line 378 in 56856c4

auto cell = TCell::Make(i);

But need to be improved for posting table with covered columns data