Revert "Reland^2 "Fix VkTestUtils's VulkanPreferredFeatures usage""

This reverts commit 0e0d6e799facb0dd637b73f926b1b1ff803c7705.

Reason for revert: Still leading to Cuttlefish errors for unclear reasons.

Original change's description:
> Reland^2 "Fix VkTestUtils's VulkanPreferredFeatures usage"
>
> This is a reland of commit 733217467faf2a2ad0f5f096019eb6d885540cd8
> * Should be fixed by restoring the SYNC_FD extension check in the tests.
>
> Original change's description:
> > Reland "Fix VkTestUtils's VulkanPreferredFeatures usage"
> >
> > This is a reland of commit 72009acc508c4ddeee63fc8a60c53a232b907258
> >
> > * Fix required including another extension (VK_KHR_EXTERNAL_SEMAPHORE_FD), re-adding protected memory feature query, and exiting backend context creation early if protected is requested but not supported (rather than an SkASSERT which stopped protected tests rather than skipping them on devices that do not support protected memory).
> >
> > Original change's description:
> > > Fix VkTestUtils's VulkanPreferredFeatures usage
> > >
> > > * This CL fixes extension mismanagement issues and includes updates for clarity and style modernization in the touched areas (verified by confirming that VK_KHR_driver_properties extension is included as expected after previously being inadvertently excluded).
> > >
> > > The issue:
> > > * This is a follow-up to the landed change "Remove redundant or outdated extension usage in Vulkan tests" (https://skia-review.git.corp.google.com/c/skia/+/1174856), which exposed an issue with our usage of VulkanPreferredFeatures where we would pass in only the list of extensions the tests want to enable. Instead, we should be providing the list of all *available* extensions on the device.
> > >
> > > * This misuse, combined with TestUtils removing extension names that should be redundant (i.e. already checked for in VulkanPreferredFeatures), led to certain key extensions/features no longer being utilized. Notably, this included VK_KHR_driver_properties. This meant driver correctness workarounds could not be enabled, which also (likely artificially) led to perf improvements.
> > >
> > > Anticipated impact of this fix:
> > > * This change is anticipated to trigger perf regressions from the current state. In reality, we should just be returning to baseline performance just before the extension removal CL landed.
> > >
> > > * Once this lands and we get perf numbers, I plan to determine whether actually not enabling certain extensions would get us performance wins without encountering driver correctness issues.
> > >
> > > Bug: b/487695539
> > > Change-Id: Idef9ac665485a50b6bd032d5138855f359f7858d
> > > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/1181116
> > > Reviewed-by: Greg Daniel <egdaniel@google.com>
> > > Commit-Queue: Nicolette Prevost <nicolettep@google.com>
> >
> > Bug: b/487695539
> > Change-Id: Ibf3bdde8bfdd46ad7018cb72331a72d43012f1ff
> > Reviewed-on: https://skia-review.googlesource.com/c/skia/+/1190458
> > Commit-Queue: Nicolette Prevost <nicolettep@google.com>
> > Auto-Submit: Nicolette Prevost <nicolettep@google.com>
> > Reviewed-by: Greg Daniel <egdaniel@google.com>
> > Commit-Queue: Greg Daniel <egdaniel@google.com>
>
> Bug: b/487695539
> Change-Id: I2758336489b75306f07b663df8270d11d2694ba9
> Reviewed-on: https://skia-review.googlesource.com/c/skia/+/1204597
> Commit-Queue: Nicolette Prevost <nicolettep@google.com>
> Reviewed-by: Noelle Scobie <nscobie@google.com>

Bug: b/487695539
No-Presubmit: true
No-Tree-Checks: true
No-Try: true
Change-Id: Ifdb03de07c310d53799ef54c3aeb9ce94938ee85
Reviewed-on: https://skia-review.googlesource.com/c/skia/+/1205056
Auto-Submit: Nicolette Prevost <nicolettep@google.com>
Commit-Queue: Noelle Scobie <nscobie@google.com>
Reviewed-by: Noelle Scobie <nscobie@google.com>