[BUG] Rounding exptimes for winter images

[robertdstein]

Describe the bug
Right now we round down winter exposure times (

mirar/mirar/pipelines/winter/load_winter_image.py

Line 131 in 97ee0d3

header["EXPTIME"] = np.rint(header["EXPTIME"])

). We also mark zero-exposure frames as bias (

mirar/mirar/pipelines/winter/load_winter_image.py

Line 83 in 97ee0d3

if header["EXPTIME"] == 0.0:

). We also take winter flats with <1s exposure time. As a consequence, J/H flats get mistagged as biases.