Investigate cases where averages are slightly higher than input data. #542
Description
🐛 Bug Report
User supplied code has highlighted a case where the area weighted average of data not exceding 1 was, in places, slightly higher than 1 (by a very minor fraction, ~1.0000000000000002). Investigate if this is baked into the ESMF weights or if there is something in the iris-esmf-regrid code which causes this, and if there is anything we can to to prevent this.