feat(chat): give every Skype emoticon a sound

Nine emoticons had audio; the other 98 fell through to a 404 and the
silent-catch, so the feature read as broken for anything but (skype),
(call) and friends — the picker is full of (smile)/(yes)/(party), none of
which made a noise.

Classic Skype never shipped per-emoticon audio (its archive is ~40 event
sounds), so there was nothing to source the rest from. scripts/gen-skype-
sounds.py synthesizes them instead: each emoticon gets a semantic
archetype — sparkle for the cheerful ones, sad for the downcast, clink for
drinks, zap for the ninjas, fanfare for the heroes — plus a pitch drawn
from its slot in that family, so a family reads as a family and no two
sounds are identical (verified: zero duplicate files). Peaks normalized to
about -3 dBFS, 0.44-0.94s, 5-11KB each.

The nine authentic Skype sounds stay as they are. Generated audio is
original work; the script is committed so the set is reproducible and
auditable, while the files themselves ship from the CDN like the rest.

Coverage verified end-to-end: all 107 catalog keys return 200 from
cdn.cumora.ai, and ten of them play in the running app.