CTS test to ensure that core features are homogenous

one of the reasons the proposals for ifuncs (https://github.com/google/android-riscv64/issues/8) are complicated is that risc-v explicitly supports the idea of different cores supporting different sets of extensions. that's obviously a nightmare for app developers, and in the arm world always ended up with kernels having to lie and just report the subset of hwcaps that were available on all cores. since this is _more_ likely to be a problem with risc-v (where it's a deliberate mistake rather than an accidental mistake), we should have a CTS test that checks for such problems. if the current vdso+syscall proposal goes in, we should be able to compare the results, and report any differences as a failure.