use MD5 hashing to avoid long index names

[sdebruyn]

resolves #317

[NicolasPA]

This index naming change is a breaking change which causes errors on our DBT models that create clustered index:
Cannot create more than one clustered index on table *. Drop the existing clustered index * before creating another. (1902)

Additionally, this would silently create duplicated indexes for non-clustered indexes because they can be more than one.

The easy is workaround is to full-refresh the DBT models, but it may be computationally costly.
The laborious but computationally efficient workaround is to manually rename all the indexes according to the new naming scheme.

Maybe it could be marked as a breaking-change in the v1.4.1 release note to help users who will get the errors.