fix: replace curl_error_map with switch to fix Static Initialization Order Fiasco

[MarkRakhmatov]

Closes #1295

[COM8]

@MarkRakhmatov thanks for contributing and fixing this! How did you debug the cause there? I also noticed this issue in the NuGet debug build but only when using msvc and I was unable identify it.

[MarkRakhmatov]

@MarkRakhmatov thanks for contributing and fixing this! How did you debug the cause there? I also noticed this issue in the NuGet debug build but only when using msvc and I was unable identify it.

You are welcome :)

I saw this issue after refactoring tests and moving them from the main function to global test suites.
I debugged code with MSVC debug build and got read access violation inside unordered_map, so I started to go over the stack to see what object this is.

I know that such strange lifetime bugs usually occur when you access object that is not initialized yet, so first assumption was Static Initialization Order Fiasco

By the way, linter was suppressed for curl_error_map so for me it also was a sign that linter tried to warn about lifetime issues.
IMO it is good practice to specify exact rule that should be skipped and good reason, and generally avoid global objects (except for simple numeric constants like constexpr size_t gBlockSize = 1024;)