Ignore lambda numbers/hashes when diffing binaries

I am analyzing a binary that is moving from C++17 to C++20 mode to see why and where it changed size. This binary has a lot of lambdas in it. Unfortunately the MSVC naming pattern for lambdas changes based on the mode so the diff is almost impossible because every lambda is showing up as only-in-base or only-in-comparison. That leaves me unable to see the true diff.

The lambda names have a hash (cpp17) or incrementing number (cpp20) in the name. These can hopefully be ignored by the comparison while still giving accurate results.

Here is an anonymized example of what I'm observing (where HASH is a long pseudo-random string)

cpp17:

winrt::Contoso::implementation::MyClass::MyFunction::_l2::<lambda_HASH>::<lambda_HASH>::operator()()const

cpp20:

winrt::Contoso::implementation::MyClass::MyFunction::_l2::<lambda_1>::<lambda_1>::operator()()const

No source files, tests, or entry points are named in the issue. Start by locating the binary comparison and symbol-name normalization code, then reproduce the anonymized MSVC C++17/C++20 lambda names; done means hash- and number-based lambda names no longer create false only-in-base or only-in-comparison results.