Fix intermittent debug failures with Debug build

[jkwak-work]

This PR replaces enable/disable style C function calls with C++ RAII style code.

In debug build, when an assertion failed in between enable and disable functions, an exception is thrown and the disable function is not called. RAII style code is safer for an exception

[jkwak-work]

With this fix,

• "windows/debug" shows <1% failure rate (0 failures out of 107 tries)
• "windows/release" shows <1% failure rate (0 failures out of 150 tries)
• "linux/release/gcc" shows <2% failure rate (2 failures out of 121 tries)
• "macos/release/clang" shows 8% failure rate (4 failures out of 50 tries).

When failures are observed on "linux/release/gcc", the following test failed.

tests/language-server/robustness-7.slang

When failures are observed on "macos/release/clang", the following tests failed,

tests/language-server/robustness-8.slang
tests/language-server/vector-member.slang

Note that without this fix, the failure rate of linux and macos wasn't 0%.
On "linux/release/gcc", the failure rate was 2%, which is about the same.
On "macos/release/clang", the failure rate was 4%, which was lower.

You can find more information here.

[jkwak-work]

closes #7333

[gtong-nv]

I agree that the RAII style is better. But I don't understand how it helps the intermittent assert...
Looks like for some autodiff code, there are some inst. that are not able to pass IR validation so we disable that check first, and then we reenable it afterwards. For example,

            disableIRValidationAtInsert();
            diffBuilder->addBackwardDerivativePrimalContextDecoration(callInst, intermediateVar);
            enableIRValidationAtInsert();

It's always disable first, and then enable again.
If something happens in between, we should in the _enableIRValidationAtInsert = false state, and no validation would be done, hence no more assert would be trigger. Seems like we are seeing the opposite.

[gtong-nv]

LGTM