Feat/a5 hooks channels

[XanderD99]

Please ensure you adhere to every item in this list

• This PR was pre-approved by the project maintainer
• I have self-reviewed the code
• I have added all necessary tests

Describe your changes

Follow-up to the A5 index (#817): A5 is both an output keyword (A5 <resolution>, like HASHES) and a search area keyword (A5 <cell>, like HASH). The output parser always won, so the cell ID was consumed as a resolution and A5 <cell> never reached the area parser:

SETCHAN chan WITHIN fleet FENCE A5 51575d8000000000
> ERR invalid argument '51575d8000000000'

INTERSECTS key A5 <cell> failed the same way — only the workaround of naming an output first (IDS A5 <cell>) worked.

Fixed in parseSearchScanBaseTokens using the same lookahead BOUNDS already uses for its identical collision: on WITHIN/INTERSECTS, if the token after A5 isn't a valid resolution (0–30), it's a cell ID, so A5 is left to the area parser. SETHOOK/SETCHAN route through the same function, so hooks and channels are covered by the one change.

Also added the A5 case to parseArea, so TEST ... A5 <cell> works alongside HASH/QUADKEY/TILE. CLIP is rejected there, since a pentagon isn't a rectangle.

Tests: channel and webhook fences over an A5 cell (enter/inside/exit/outside, plus "a5" in the message when the output form is used), SETCHAN/SETHOOK parsing with and without an output, CHANS/HOOKS round-trip, error cases, an AOF-replay test since the stored fence command is re-parsed on restart, and TEST coverage matching the existing QUADKEY cases. All fail without the changes.

Issue number and link

#817 — #817

[XanderD99]

Good call — reworked to use the existing format convention instead of guessing.

A5S <zoom> is now the return format, alongside POINTS/IDS/HASHES/OBJECTS, and A5 <cell> is only ever a search area. The parser no longer has to decide whether the token after A5 is a resolution or a cell id, so WITHIN/INTERSECTS and the hooks and channels they back all read the same way:

INTERSECTS fleet A5S 10 A5 51575d8000000000
SETCHAN c INTERSECTS fleet FENCE A5S 10 A5 51575d8000000000

GET key id A5 <zoom> and SET key id A5 <cell> stay as they were — those mirror HASH and never collided.

Two things fell out while testing it:

• The A5 output was never wired into the JSON envelope, so SCAN key A5 10 with OUTPUT json returned malformed JSON ({"ok":true{"id":"a",...}]) and silently dropped WITHFIELDS fields. It now emits "a5s":[...] like every other output.
• A5 1 panicked the server. 1 parses as hex but carries no resolution marker, so GetResolution returns -1 and a5-go reaches SToAnchor with resolution -2 → makeslice: cap out of range. Reachable from any client through INTERSECTS, WITHIN, TEST or SET. a5DecodeCell now validates the resolution and origin bits, with a sweep test asserting nothing it accepts can panic the library.