restart indexing pipeline on index update

[trinity-1686a]

Description

restart indexing pipelines on index update
part of #5084

How was this PR tested?

todo

[github-actions]

On SSD:

Average search latency is 0.999x that of the reference (lower is better).
Ref run id: 3206, ref commit: 9d8fab0
Link

On GCS:

Average search latency is 1.06x that of the reference (lower is better).
Ref run id: 3207, ref commit: 9d8fab0
Link

[trinity-1686a]

this still needs automated tests, but should otherwise be ready for a first round of review

[rdettai]

It would be great to add a test this in https://github.com/quickwit-oss/quickwit/blob/main/quickwit/quickwit-integration-tests/src/tests/update_tests/doc_mapping_tests.rs

[trinity-1686a]

thinking about it, as we only restart indexing pipelines, but don't do anything about shards, shards from ingest-v2 could have a stall doc_mapping, and if validation is enabled, they could reject documents that are valid under the new mapping. Ideally we'd want to also close all shards for the index and open as many new ones.
We would then end up with a race condition where we said "this document is valid" at validation, run it against the new mapper, and throw it away as invalid then, without notifying the writer.