Odd drawing behavior with radicals and MolsToGridImage

[pstjohn]

Occurs with rdkit version 2021.03.02, (and 2021.03.01, but this seems to be a recent change. I could try to track down an older rdkit version that didn't have this issue if that's helpful)

https://gist.github.com/pstjohn/d7887e6349009e414bd84a18fbc6b952

Including a radical in a molecule for MolsToGridImage seems to affect the scaling considerably, but oddly only if it's not in the first position.

It doesn't happen with all radicals though, smaller ones don't trigger the rescaling.

[DavidACosgrove]

Groan. It’s my least favourite part of the drawing code. I’ll take a look tomorrow. Thanks for the clear and concise report.
Dave