Allow tolerance in tests.

[plenaerts]

Hi, in debian QA tests, we see an error on riscv64 architecture.
See https://ci.debian.net/packages/r/r-bioc-s4vectors/testing/riscv64/69538974/#L1763 for the error.

I propose the attached patch to fix this. Thanks for taking it in consideration.

Br,

Pieter

[hpages]

Thanks for the PR. Any reason not to use checkEqualsNumeric() like in the other checks where some tolerance is needed? I'd rather keep some consistence across the checks if possible.

[plenaerts]

You're right. There is no good reason. I applied your suggestion. Should've used the more specific function in the first place. 🫣

[hpages]

Thanks. Can you confirm that the default value (.Machine$double.eps^0.5) for the tolerance argument is too strigent on the riscv64 arch, hence why you're explicitely setting the argument to 1e-6? Thanks again.