bytes package tests test implementation defined behaviour

[mwhudson]

As can currently be seen at https://people.canonical.com/~ubuntu-archive/proposed-migration/update_excuses.html#golang-github-labstack-gommon, the tests for the bytes package fail on arm64, ppc64el and s390x architectures. This turns out to be because the tests depend on the conversion of the (just) too large float math.Pow(2, 64) to int64 returning MinInt64. On these platforms it seems the conversion saturates and returns MaxInt64. FWIW the standard says:

In all non-constant conversions involving floating-point or complex values, if the result type cannot represent the value the conversion succeeds but the result value is implementation-dependent.

so the tests shouldn't depend on this, or if this behaviour is important then the implementation needs to take care to enforce it. I don't really see a reason why the tests need to push the edge of the range of int64 like this so I'll probably patch them in Ubuntu to test smaller values.