aset/aref for char-tables, range-tables etc.

i originally decided that it was a bad idea to allow aset/aref to operate over char-tables and range-tables, like FSF does. but now i'm not so sure. for one, i recently added a change to use char-tables in place of 256-element vectors for the return value of `make-display-table', and code in w3 and x-symbol (and probably elsewhere) broke. but more to the point, i'm influenced by python, which uses [] for accessing arrays and hash tables ("dictionaries"), including ranges in arrays (the slice notation); and both char-tables and range-tables are basically just sparse arrays, with some optimizations. comments? stephen?