cryfs 0.11.3-5 fails to build on ppc64el

cryfs 0.11.3-5 fails to build on ppc64el due to some test cases failing with:

```
[ RUN ] Encrypted_AES256_CFB/BlockStore2Test/0.NumBlocksIsCorrectOnEmptyBlockstore
unknown file: Failure
C++ exception with description "Error calling mlock. Errno: 12" thrown in the test body.
[ FAILED ] Encrypted_AES256_CFB/BlockStore2Test/0.NumBlocksIsCorrectOnEmptyBlockstore, where TypeParam = EncryptedBlockStore2TestFixture<cpputils::AES256_CFB> (0 ms)
```

errno 12 is ENOMEM which can have multiple reasons.

Discussion on #launchpad on 2023-05-16:

<cjwatson> bdrung: I doubt it's a question of physical memory (they get 16GB). From the log it looks more like mlock
<cjwatson> IIRC upgrading the builders to jammy will help with that
<cjwatson> (due to systemd changes)
<bdrung> can you elaborate your reasoning what the problem with mlock is and why jammy could help?
<cjwatson> The reason I suspect mlock is that the build log mentions it. And mlock failing is usually not helped by increasing available system memory - mlock has a much more restricted available pool.
<cjwatson> The systemd thing is a vague memory though.
<cjwatson> Possibly wrong. systemd 852b62507b22c0a986032a2c9fa9cc464a5b7bd2 looks relevant, but that's in 251 which was post-jammy.
<cjwatson> But in any case, if a test is showing that insufficient lockable memory is available, that's probably something that should be fixed in the distro - not something we should hack around specifically on builders, since that would just make the test environment less realistic.
<cjwatson> If it's just a matter of keeping up with current distro configuration, we can do that (hence my comment about jammy), but I don't think we should have something like a tweak for LimitMEMLOCK= just for builders.