Split functions to avoid 'is_manifest' control coupling.

[BillyONeal]

Extracted from #1210

[ras0219-msft]

I wonder if this pattern will be common enough to warrant a

LocalizedString LocalizedString::error_in_path(StringView path) {
    return LocalizedString::from_raw(path).append_raw(": ").append(msgErrorMessage);
}

[BillyONeal]

I think this would be worth it but I'd rather do it as its own PR rather than expanding the scope of this one

[ras0219-msft]

I feel like this function should be able to be written with less nesting, perhaps along the lines of:

std::error_code ec_manifest, ec_control;
auto content_manifest = fs.read_contents(path_manifest, ec_manifest);
const auto exists_manifest = !ec_manifest;
const auto failed_manifest = ec_manifest && ec_manifest != std::errc::no_such_file_or_directory;
auto content_control = fs.read_contents(path_control, ec_control);
const auto exists_control = !ec_control;
const auto failed_control = ec_control && ec_control != std::errc::no_such_file_or_directory;
if (failed_manifest) {
  // msgFailedToAccessManifest(path_manifest)
} else if (failed_control) {
  // msgFailedToAccessManifest(path_control)
} else if (exists_manifest && exists_control) {
  // msgManifestConflict(port_directory)
} else if (exists_manifest) {
  // success on manifest
} else if (exists_control) {
  // success on control
} else {
  // msgPortMissingManifest2(port_directory)
}

[BillyONeal]

I don't like this transformation as it does a read_contents rather than an exists check on the CONTROL in the case we know loading the json succeeded. I inverted the ec tests and that flattened the result substantially, does that resolve your hearburn?