test: Data race detected when running rpc_blockchain.py

[hebasto]

https://cirrus-ci.com/task/6068765160898560?command=ci#L5518

[maflcko]

I can't make any sense of those data races. Now that we implement our own double lock detection in #19337, we should revert the broken clang tsan (#19041) and potentially provide instructions how a non-broken tsan can be selected on systems that ship with the clang tsan by default (fedora, maybe freebsd?, ...)

[maflcko]

Actually, I could (intermittently) reproduce on libc++ and libstdc++