Inconsistency in parsing @SQ headers with multiple LN: and SN: tags

[daviesrob]

As noted in #1863, when reading @SQ headers sam_hdr_create() records the value for the last LN: and SN: tags, while refs_from_header() uses the result from sam_hrecs_find_key() which is the first one. This leads to a possible discrepancy in how such lines are interpreted by the SAM- and CRAM- reading code.

As such lines are invalid according to the specification ("Within each (non-@CO) header line, no field tag may appear more than once...") it might be best to reject such inputs by enforcing the one-tag rule. Hopefully the number of files that would be rejected as a result will be very small, if any, and it will prevent any possibility of inconsistent interpretation.