Fix output of labels and references that contain characters with an UCS4

Fix output of labels and references that contain characters with an UCS4
codepoint > 255.
I decided that this is no file format change: We already had the change to
format 249 that allowed unicode in .lyx files. The output to .tex of
non-ascii characters is now different than before (even for those that were
supported previously, e.g. german umlauts in latin1), but this is only
relevant if people referenced a label in ERT. Since we cannot detect this
anyway we don't need a file format change.

	* src/support/lstrings.C
	(escape): Extend the escaping algorithm from 8 bit to 24 bit.

	* src/support/lstrings.h
	(escape): Update comment

	* lib/lyx2lyx/lyx_1_4.py
	(lyx_support_escape): Update comment

git-svn-id: svn://svn.lyx.org/lyx/lyx-devel/trunk@15883 a592a061-630c-0410-9148-cb99ea01b6c8