inaccurate explicit/implicit hydrogen handling in fingerprints

Using the 'sub' fingerprint for substructure search prefiltering produces false negatives where the hydrogen count is specified in the SMARTS query.

Examples:

Query:
`[*:1]-[OHX2]`

Targets:
`N1C[C@@H](O)CO1.Cl`
`C(F)(F)(F)C(CN)O.Cl`

Query:
`[*:1]-[NHX3]-[*:2]`

Target:
`C(NCC)(C(C)(C)O)=O`

Performing the SMARTS matching produces hits in all cases. The prefiltering works after adding explicit hydrogens to the targets. I suggest to neglect `[H]` in the SMARTS during the sub fingerprint generation as `[H]` includes the implicit hydrogens as well according to the Daylight spec.