In the vast majority of cases where we call GRPC_ERROR_CREATE_FROM_COPIED_STRING(), we are constructing a C++ string, and then calling its c_str() method to create the error. This API is not very ergonomic, because it requires calling c_str() all over the place. It may also be slightly less efficient, since it requires reallocating the string and copying it.

To address this, I have added a GRPC_ERROR_CREATE_FROM_CPP_STRING() method, which takes ownership of the C++ string that you pass to it. Under the hood, this uses grpc_slice_from_cpp_string(), which I added back in #23157. (Note that this still makes a second allocation for the slice ref-count, so it winds up holding two different memory allocations for the lifetime of the slice, but there is no time during which the string is double-allocated, and it does not need to make a copy.)

I have also added a GRPC_ERROR_CREATE_FROM_VECTOR_AND_CPP_STRING() function to address the less pervasive but still common use-case where we want to use GRPC_ERROR_CREATE_FROM_VECTOR() but the string is a dynamically generated std::string instead of a static string.