Restored benchmark collector job spawn after fixes done in MR: benchmark-results-collector!7

build.sh script additionally immediately uses atomic package to enforce its final build on builder image. Then runtime image can be smaller due to eliminated compiler.

Old upgrade code (specific to previous releases) removed and new upgrade procedure adjusted to HAF requirements.

hivemind_app_block_view switched to hive.blocks_view due to peformance reasons. Postgres again is ignoring index, while perfoming order by query over indexed column, in case of complex view.

Updated haf submodule to the version containing required indexes and functions to wait until HAF instance will be ready for processing.

Mateusz Żebrak authored Nov 10, 2022 and Bartek Wrona committed Dec 16, 2022

Due to performance issues when querying
for last num of `app_name.blocks_view`,
additional function get_head_state()
was created.

hive.hivemind_app_blocks_view data are provided by simple subquery instead of JOIN, since at least PostgreSQL 12 planner misbehaves and uses table scan instead of index scan to select row(s) from hive.blocks table, what kills performance.