NSCFString: fix lengthOfBytesUsingEncoding recursion and getCString encoding direction

[DTW-Thalion]

Summary

Two bugs in the bridged-CFString (NSCFString) encoding methods:

1. lengthOfBytesUsingEncoding: called itself unconditionally:

   - (NSUInteger) lengthOfBytesUsingEncoding: (NSStringEncoding) encoding
   {
     return [self lengthOfBytesUsingEncoding: encoding];   // infinite recursion
   }

   This recurses until the stack overflows. Implemented by converting the
   string with CFStringGetBytes and returning the number of bytes used.

2. getCString:maxLength:encoding: converted its NSStringEncoding
   argument with CFStringConvertEncodingToNSStringEncoding (the CF→NS
   direction) before passing it to CFStringGetCString, which expects a
   CFStringEncoding. Use CFStringConvertNSStringEncodingToEncoding, the same
   direction the neighbouring methods (initWithBytes:,
   maximumLengthOfBytesUsingEncoding:, …) already use.

Test

Tests/CFString/bridge_encoding.m builds a bridged CFString and checks
lengthOfBytesUsingEncoding: (which previously crashed) and
getCString:maxLength:encoding:.

Note

dataUsingEncoding:allowLossyConversion: has the same reversed conversion, but
it cannot currently be exercised end-to-end because
CFStringCreateExternalRepresentation returns NULL for these encodings — a
separate issue, so it is left out of this change.