Better handling of conversion to/from Tcl Strings

[rkeene]

Right now we use duk_safe_to_lstring() (and duk_push_lstring()) a lot which converts an object to a CESU-8 encoded string which we can consume with Tcl_NewStringObj() (which expects UTF-8 string) because the only differences between CESU-8 and UTF-8 are outside the Unicode Basic Multilingual Plane (BMP).

However:

• Tcl can be compiled with greater Unicode plane set support
• Newer versions of Tcl will support a larger Unicode plane set

There should be some mechanism to deal with this added in the future, for emoji !