Reject a blank row in the gencost matrix instead of reading past the string

[arpitjain099]

readMatPowerGenCostRow checks the trailing semicolon with row[row.size() - 1] after rtrim. On a blank line the string is empty, size() - 1 wraps, and the subscript reads past the buffer. The gencost loop in readMatPower hands every line up to ]; to the row reader without filtering empty ones, so a case file with a stray blank line between two gencost rows reaches it.

Built with -fsanitize=address against the header as it stands:

control (no blank line): parsed, rows=2
ERROR: AddressSanitizer: stack-buffer-underflow
  #0 readMatPowerGenCostRow MatpowerParser.hpp:156
  #1 readMatPower MatpowerParser.hpp:282

The file that triggers it is a normal case file, nothing malicious, just an extra newline in the matrix.

The condition already throws matlab_syntax_error for a row without a trailing semicolon, so adding row.empty() to it puts a blank row on that same path rather than inventing new behaviour. I added the case to test_parse_gencost_row, which aborts under ASan before the change and passes after it.

One thing I noticed while checking the siblings, not changed here: readMatPowerGenRow and readMatPowerBranchRow both end with checkEndOfMatrixRow, so they reject a blank row, but readMatPowerBusRow does not call it. A blank line in mpc.bus is currently accepted and appends a bus row with every field zero. Happy to send that separately if you want it tightened up, since it changes what the bus parser accepts rather than fixing a read.

[pelesh]

@arpitjain099, thanks for your contribution. Please create an issue with reproducer and suggested solution. Then we can decide how to move forward. I will close this PR until then.