Bomb planted — *where*: the message line and the kill feed name the place, in the map's own words #174

Use the vocabulary #167 adds to tell the person where something happened, so a plant you did not see is a place you can run to rather than a beep you have to hunt for.

Today say() produces Bomb planted and nothing more: the site is on the radar as a letter, and the beep tells you the rest. Once placeAt(map, pos) exists, the same information can be a word. Small, and it makes the places work pay for the person as well as for the bots.

Waits on #167 (and therefore on pull #172, which carries it).

Done looks like

• [ ] The plant message names the place the bomb is actually standing in — Bomb planted — B site — resolved with placeAt and nothing else. No second vocabulary, no table of coordinates.
• [ ] The defuse message and the one for a bomb dropped in the open get the same treatment, or say why they do not.
• [ ] It is one lookup at the moment of the plant, not per tick: world.ts is the arena's simulation too.
• [ ] Nothing changes for the bots: this is the message line the HUD draws, not observe.ts, not the policy's 32 inputs, and not what plan.ts decides.
• [ ] Tests: a bomb planted on each site on each map says that site's name; the message is produced once per plant; a round restart clears it.
• [ ] README: the plant's line under Playing says it names the place.

Files I expect to change: src/client/world.ts, src/client/world.test.ts, README.md.

Risks: if #172 is closed rather than merged, this waits on #167 being built on its own instead. Keep the wording short — the HUD holds a message line for a few seconds and a long sentence is a line nobody finishes reading.

Brief that fits: scout.