Allow filter by VariantId using SampleIndex

[j-coll]

The fastest way of filtering by a small set of variants is going directly to the HBase table. This assumption was made initially to discard/postpone the variantId filtering using the SampleIndex.
This assumption is not that true when you're filtering also by other positional filters, like Region or Gene (which is, ultimately, a region filter as well). In this case, the query won't use the SampleIndex, but the SecondaryIndex (Solr). This approach is still correct, but slower, and with a poor approximated count.