The {N$$sep$$} brace prefix now also accepts a count range ({N-M$$}) and a
~ independent flag ({N~$$} / {N-M~$$}, repeats allowed; absent = unique).
Both tokenizers (Python engine + TS richTokenize) parse it identically
into dp-multi meta {min, max, independent}; legacy `count` is kept (== max)
so existing readers + the resolver behave unchanged for a fixed {N$$}.
Shared syntax-corpus.json locks Py≡TS parity (existing multi_pick entries
gain the new keys; new range/lone-ref cases added).
Resolution of the range/independent semantics lands next (engine resolver);
this commit is tokenizer + parity only — a fixed {N$$} is byte-for-byte
unchanged.
DumiFlex