Check again m/z deviation from Spectra <-> MsmsSpectrum

[tnaake]

When converting between Spectra and spectrum_utils's MsmsSpectrum, the m/z values are not identical (deviation approx 1e-06). Look into the nature of this problem, see also PR #47.

[bittremieux]

So spectrum_utils represents m/z values as double precision floats (float64), whereas in R the default is single precision (float32). I hypothesize roundtripping from R to Python back to R (i.e. float32 -> float64 -> float32) results in some rounding errors. Differences of 1e-6 sounds about right for that.