#2575 serve: the READY timeout's message names STRATA_ENGINE_READY_S (#2331)

Description

The code part of #2331 (the reporter's own diagnosis in the thread: not an engine exit, the 900 s READY limit).

**What it changes.** `serve/server.py` ends an engine that has not reported READY within `ENGINE_READY_S` (900 s by
default, `STRATA_ENGINE_READY_S`). On the reporter's box the start needed longer (`--ple-io ram` alone read the
28.8 GB PLE table in 454-508 s from HDD-class storage, then the expert RAM copy), every run ended at ~15 min 07 s,
and the message ("the engine did not report READY within 900 s") did not name the setting, so it looked like a
silent engine exit. The message now ends with "(STRATA_ENGINE_READY_S; a load from a slow disk can take longer: set
it higher, or 0 to wait for ever, before the server starts)". The variable is read when the server starts.

**What it leaves alone.** The limit and its default, every other start failure message, and the docs (a row for
`STRATA_ENGINE_READY_S` in docs/DETAILS.md is a separate docs change; it is not listed there today).

**How it was checked.** `serve/test_server.py` `UntimedReads.test_engine_that_never_says_ready_is_ended` now also
checks the variable's name in the error (fails on `main`, passes here). With the venv's Python: `serve.test_server`,
`serve.test_lifecycle`, `serve.test_fatal_recovery`, `serve.test_ready_timeout`: OK. Python 3.14, Ubuntu (Linux 7.0).

🤖 Generated with [Claude Code](https://claude.com/claude-code)