I don't know if such a tool or pipeline exists off-the-shelf. If not, I think some scripting is required, probably using python/pysam java/htsjdk or other API to process bam files.

Essentially for each read spanning a variant and spanning one or more cytosines on the reference sequence you record whether the cytosine(s) are methylated or not (i.e. they are C or they have been converted to T) and you record whether the variant is a reference or an alternative allele. You need to record also the position of the cytosine(s) in genomic coordinates. (This strategy is not difficult but it might be a bit tricky to get right.)