Experiment 5, testing "stable" channel by that name too

It is 1.84 currently, but only one of the 1.84 jobs failed in
experiment 4 (at least the first full run of that experiment), and
the failure occurred in the runner software itself this time, while
1.83 had a failure in `actions/checkout` and a SIGSEGV in `rustc`.
This is shown in the logs of the first workflow run at:
https://github.com/EliahKagan/gitoxide/actions/runs/12964283126

This commit tests the stable channel by name as well as by version
number in case that somehow makes a difference (though it probably
can't).

.github/workflows/arm-segv-experiment.yml

@@ -7,8 +7,8 @@ jobs:
     strategy:
       matrix:
         os-ver: [ '22.04', '24.04' ]
-        toolchain: [ '1.81', '1.82', '1.83', '1.84' ]
-        number: [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31 ]
+        toolchain: [ '1.81', '1.82', '1.83', '1.84', 'stable' ]
+        number: [ 0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 ]
 
       fail-fast: false