The retime or synchronize would likely be more apporpriate here, however the resample function could work (and interp1 is an option as well). I would downsample the hourly data to the every-six-hour data rather than upsample the every-six-hour data. The reason is that upsampling creates data where none previously existed. You cannot determine what the IFREMER data actually were at the hourly times, so any comparisons of interpolated hourly data to the ERA5 data would likely not be valid.
So I would use retime or synchronize to downsample the ERA5 data to the same times as the IFREMER data and go from there.
