It wouldn't damage the files, but it wouldn't add support for those characters. When Decker loads a .deck file (including one embedded in a .html export), it normalizes UTF-8 into the supported DeckRoman character set and converts unsupported characters into �. A manually-edited deck would lose unsupported characters before scripts have a chance to run, and the file would not round-trip retaining that data.
As in Native-Decker, it would be possible to create a modified version of Web-Decker with a different set of lookup tables for converting between UTF-8 and DeckRoman. (This would make such decks incompatible with mainline Decker!) You'd also need to furnish a set of fonts corresponding to this rearranged text encoding- presumably at least replacing Decker's three built-in fonts 'Mono', 'Menu', and 'Body'.
If you just want to display text in a non-latin alphabet you can simply make custom fonts using Decker's font editor, but the underlying text will still be DeckRoman from the perspective of copying and pasting to the system clipboard, typing on a keyboard, or saving a .deck file to disk.