Remove from grpauto.tst a test which exceeds memory limits

[olexandr-konovalov]

This will allow to avoid test failure reported in #2377 in
stable-4.9 branch. The fix in the master branch will be done
differently. This test is duplicated in benchmarks, so we will
monitor its status there.

[hulpke]

For stable this is fine. In master the problem will go away once #2383 is merged

[hulpke]

Just curiosity: How did we decide on the 1G limit? These larger group theoretic calculations are testing a lot of different (and involved) functions on the way.

[olexandr-konovalov]

@hulpke -o 1g was used in make test... targets for quite long time - I can't find a link to some discussion about possibly raising limits. IIRC last time we still decided to keep -o 1g which is lower than default. My argument in favour of 1g is that it helps to track memory regressions - like in this case, which would remain unnoticed in default settings.

Today, if you want to add a genuine test which demands more memory, I suggest to put it to benchmark, which allow to use different (also standardised) memory limits (this is the current setup, open for suggestions and improvements). Benchmarks are run weekly on Jenkins CI in St Andrews.