[ML] CStringStoreTest.testStringStore failure

[davidkyle]

[Exception] - critical check strings.size() == CStringStore::names().m_Strings.size() has failed [16 != 14]
 == [File] - CStringStoreTest.cc
 == [Line] -165

Failed on macOS aarch64 only

https://elasticsearch-ci.elastic.co/job/elastic+machine-learning+pull-request+macos-aarch64/624/

[davidkyle]

Another failure this time on Windows

[Exception] - critical check uniques.size() < 20000 has failed [22668 >= 20000]
 == [File] - CStringStoreTest.cc
 == [Line] -213

https://elasticsearch-ci.elastic.co/job/elastic+machine-learning+pull-request+windows/697/testReport/junit/lib.model.CStringStoreTest/CStringStoreTest/testStringStore/

[droberts195]

I just noticed that gcc has a -fsanitize=thread option. Maybe this could be used to check if the problem is an actual race condition. If it's not then the failures just mean that lots of clashes are occurring during some test runs, causing multiple CStoredStringPtr objects to be created which reference the same string, and then the test expectations simply need updating to reflect that.

I also opened #2130, because maybe the best way to fix this once and for all is to completely remove CStringStore from the codebase - this needs further investigation - there may still be a reason why we need it.