hoaparse: also accept LBTT input
This is probably the worse grammar I wrote: the LBTT format is designed to be scanned with scanf, and very inconvenient to parse with bison/flex. Here the scanner basically has to emulate a parser to classify the different INTs as tokens with different types. * src/hoaparse/hoaparse.yy, src/hoaparse/hoascan.ll: Add rules for LBTT. * src/hoaparse/parsedecl.hh: Add a way to reset the parser between each automata. * src/tgbatest/hoaparse.test, src/tgbatest/lbttparse.test: Add more tests.