Force warnings on in the test runner

We should force warnings on in the test runner (at least `-Wall -Wextra -Werror`, maybe even `-Weverything` with Clang). We want to catch any issues in our code so we don't prevent users from using these flags.

Start by locating the test runner and reviewing how compiler flags are configured for its builds. Check the runner with GCC and Clang using -Wall, -Wextra, and -Werror, then assess whether -Weverything is viable. Done means the runner enforces the agreed warning flags and the existing tests still pass.

Written by the indexing model from the issue text.