Exhaustive enum switch

[cjen1-msft]

This adds -Wswitch-enum which requires a branch for every type of an enum.
And it also adds -Wcovered-switch-default which throws an error if every case of an enum is covered.

And then fixes all of the switch statements.

[achamayou]

Suggested change

	throw std::logic_error(fmt::format("Unknown curve {}", curve_id));
	throw std::logic_error(fmt::format("Unsupported curve {}", curve_id));

[achamayou]

Suggested change

	throw std::logic_error(fmt::format("Unknown JWK curve {}", jwk_curve));
	throw std::logic_error(fmt::format("Unsupported JWK curve {}", jwk_curve));

[achamayou]

These are ECDSA, not EdDSA, so ideally we should find a way to validate them separately. The message would be something like "Not a valid EdDSA curve then".

[achamayou]

Suggested change

	throw std::logic_error("Unhandled value for ForwardingRequired");
	throw std::logic_error(fmt::format("Unhandled value for ForwardingRequired: {}", v));

Slightly more debuggable.

[achamayou]

Suggested change

	throw std::logic_error("Unsupported SEV-SNP product for TCB version policy");
	throw std::logic_error(fmt::format("Unsupported SEV-SNP product for TCB version policy: {}", product));

[achamayou]

I am confused as to why there is no default here, but we have it in other places (like to_string(Product) below) where we are ostensibly exhaustive.

[achamayou]

Suggested change

	throw std::logic_error("Unknown SEV-SNP product");
	throw std::logic_error(fmt::format("Unknown SEV-SNP product: {}", product));

[eddyashton]

Is this supposed to be unreachable? I understand a default-throw, that's "this might currently be unreachable, but if we make it reachable in future it throws". But this trailing throw, if we now think we have compile-time enforcement that it remains unreachable? I think that should be std::unreachable() (or nonstd::unreachable() calling __builtin_unreachable(), until we have C++23 support).