YT-21084, YT-21194: Secondary index modification refactoring

Addressed a concern regarding performance - a hash-map used to be built for every row in transaction which was hardly optimal.
New pipeline allows us to build a resulting table row according to a predetermined mapping, which will be useful for predicates in indices.
The use of a single name table makes the code more clear with less implied invariants.
Also fixed
    - a bug with doubling rows in index table and added a test;
    - a use-after-free bug with looked up rows.
    - sequencing bug when using rpc proxy.
6f3f8915e7d9952193897acffea5cdbdbe1201d7

0 files changed, 0 insertions, 0 deletions