vmm_tests: Fix build requirements function when parsing vmm_test macros

[damanm24]

The build_requirements function did not take into account all the possible combinations for a VmmBackend to be configured. This allowed for some discrepancy when generating the test case host capabilities requirements depending on how the vmm backend was being configured. This PR refactors the make_vmm_test so that resolving the vmm backend happens earlier in the function, and the resolved value is the one that is used throughout.

[Copilot]

Pull Request Overview

This PR fixes a bug in the VMM test macro where build requirements were not correctly accounting for all possible VMM backend configurations. The fix resolves the VMM backend (HyperV or OpenVMM) earlier in the process and uses that resolved value consistently throughout the test generation.

Key Changes

• VMM backend resolution now happens at the start of the config loop, combining specific_vmm and config.vmm into a single resolved_vmm value
• The build_requirements function signature changed to accept the resolved VMM backend directly instead of the entire config
• Duplicate VMM resolution logic in the match statement for cfg_conditions was removed

[github-actions]

At least one Petri test failed.

[smalis-msft]

Just spitballing here, feel free to say no if you like the match better

if specific_vmm == config.vmm || specific_vmm.is_some() != config.vmm.is_some() {
    specific_vmm.or(config.vmm).ok_or(Error::new(config.span, "vmm must be specified"))?
}
else {
    return Err(Error::new(config.span, "vmm mismatch"))
}