chore: Improve general SLO testing

[Zork33]

1. Make parallel testing of several PRs in different frameworks possible. For this purpose switch from rigid cluster locking for the time of testing execution to a counter of simultaneously running SLO tests.

2. Make parallel launch of old and new versions of code and display the difference on the graph. The new version is the last commit in PR. The old version is the commit where the PR branch and the main branch are separated

It is expected that the result will be informative even in case of cluster degradation when running SLO tests of multiple frameworks at the same time, since there will be a comparison between the old and new version of the code

Expected difficulties: Since currently all work with the cluster is done within the actions of a single YDB SDK, there are likely to be questions:

1. How to collectively lock the cluster with ref-counter and
2. How to continue to "pump" the cluster when tests of the framework that "pumped" the cluster are completed, and others tests are still in progress

[Zork33]

In the process of customizing my SLO_TESTS project clone so that I can locally develop a code branch with new logic using the existing product SLO tests cluster