cph-by-chenkx 2.4.2
====================

Fixed - a runtime error was reported as TLE when the source lived in a folder
whose name contains a space or non-ASCII characters (for example
"...\2026 CSP-S 复赛赛前模拟赛 day7\"). The crash replay handed gdb a quoted
program path; gdb treats those quotes as part of the name and then refuses to
load anything, so the replay found no crash, the panel fell back to the verdict
it already had, and a program that clearly aborts came out as TLE with an empty
crash line.

The path is now passed through verbatim - argv already carries the boundaries -
so the replay locates the frame in the source under test as intended.

Why this went unnoticed: every test fixture for the crash replay lived under the
system temp directory, whose name has neither spaces nor non-ASCII characters,
so the quoted branch was never exercised. The crash-line harness now also runs
the same case from a directory shaped like a real one, and the CI suite asserts
the program argument is never quoted.

Notes:

* tests/run_tests.py: 327 -> 328 checks.
* .workbuddy/harness_crash_line.py: 17 -> 22 checks (the spaced-path case
  reproduces the exact symptom - TLE with an empty crash line - against the
  previous code and passes against this one).
* Unchanged: a release build still cannot report a runtime error (silent
  undefined behaviour exits 0), and a machine without gdb still degrades to the
  libstdc++ header line instead of failing.