Deprecate vc and builtin_vc build options

[guitargeek]

The vc flag enabled two features:

1. Some vectorized Math functions, but they didn't get adopted much by our users because you need to use specific SIMD types.
2. A vectorized backend for the VecCore-powered TFormula implementation, but this doesn't pass the unit tests.

Two drawbacks of removing it:

1. If the user go through the effort of using the SIMD types for vectorization, we can expect them to implement the required math functions themselves (ROOT didn't provide many any way, VectorizedTMath.cxx had only 250 lines).
2. In the cases where the VecCore/Vc backend for TFormula actually worked and the user enabled vectorization via the TFormula constructor argument, there might be a performance regression.

The upside of removing it:

1. We remove code that is untested in the CI because enabled it makes tests fail (see this CI test run). Leaving it might harm users with silently wrong results.
2. Get rid on one (builtin) dependency.
3. Reduced build option matrix.
4. We don't need to worry about how/if builtin Vc headers should be available at install time.
5. Vc is in maintainance mode and no longer actively developed (see it's README)

Closes also the following issues:

• https://its.cern.ch/jira/browse/ROOT-10614 - TFormula tests fail when vectorization is enabled
• Vc.pcm not found when running root after building with external Vc with runtime_cxxmodules ON #18949

[github-actions]

Test Results

    21 files      21 suites   3d 20h 9m 0s ⏱️
 3 689 tests  3 683 ✅ 0 💤  6 ❌
75 632 runs  75 622 ✅ 0 💤 10 ❌

For more details on these failures, see this check.

Results for commit 57c38f5.

♻️ This comment has been updated with latest results.

[guitargeek]

The failure on alma8 is unrelated: it is a artifact from upgrading Python on that platform and doing incremental builds, so it's still looking for the Python3.8 development files that are gone.