Test shift ops in OpSpecConstantOp with more bit widths

Test shift ops in OpSpecConstantOp with more bit widths

This commit adds tests to check bit shifting operations work properly
when used with OpSpecConstantOp and different bit widths. New tests are
divided in two categories: either both the value being shifted and the
shift amount use the new bit width, or only the shift amount uses the
new bit width.

In addition, the OpSpecConstantOp test mini-framework for testing two
integers has been expanded and made more flexible in order to accomodate
the new tests and to avoid some hackish handmade conversions from
floating point to integers when passing floating point values.

Also, arithmetic bit shifting operations to the right have been modified
to use negative values in order to verify the sign bit is preserved
properly.

New tests:
dEQP-VK.spirv_assembly.instruction.*.opspecconstantop.*i64*
dEQP-VK.spirv_assembly.instruction.*.opspecconstantop.*i16*
dEQP-VK.spirv_assembly.instruction.*.opspecconstantop.*i8*

Affected tests:
dEQP-VK.spirv_assembly.instruction.*.opspecconstantop.*

Components: Framework, Vulkan
VK-GL-CTS issue: 2164

Change-Id: I3b71431ceb0022c661abdc645a2de09e9867e8ab