Extremely slow Clang + Valgrind on macOS x86_64

[hebasto]

The "x86_64: macOS Ventura, Valgrind (clang)" CI job takes more than 40 min:

• https://github.com/bitcoin-core/secp256k1/actions/runs/18495253559/job/52697859721
• https://github.com/bitcoin-core/secp256k1/actions/runs/18507649867/job/52740457457

However, "x86_64: macOS Ventura, Valgrind (gcc)" takes less than 20 min:

• https://github.com/bitcoin-core/secp256k1/actions/runs/18495253559/job/52697859732
• https://github.com/bitcoin-core/secp256k1/actions/runs/18507649867/job/52740457434

[real-or-random]

Hm, interesting observation. But for now, I'd assume it's legitimate? Perhaps, for some reason, clang's output has much more of what slows down Valgrind (e.g., direct memory accesses instead of calls to intercepted functions such as memcpy).

I'm not convinced that we should do anything about it right now. This will be resolved by #1755. If it turns out to be an issue on future Valgrind macOS jobs, we could still check which of the tests are slow. The new framework (#1734) makes this much easier.

[real-or-random]

Unfortunately, #1759 hasn't improved this.

clang, slow: https://github.com/bitcoin-core/secp256k1/actions/runs/18569028464/job/52937855454?pr=1759
gcc, fast: https://github.com/bitcoin-core/secp256k1/actions/runs/18569028464/job/52937855452?pr=1759