Fix variant region for ins and dup on intron-exon boundary

[b0d0nne11]

Fixes #655

This PR is offered as an alternative to https://github.com/biocommons/hgvs/pull/709. It solves the same problem, that certain ins or dup variants that occur at the intron/exon boundaries have non-zero offsets but are still coding. The approach taken here varies from the earlier PR in a few ways:

• It avoids overriding any methods
• It keeps most of the replace reference changes to the variant mapper
• It avoids the intermediate mapping operations
• It keeps the new tests to a single new issues test file