sTGC Strip Trigger: Fixed the scattered point segment finding inefficiencies

There are some scattered point inefficiencies for the strip trigger performance, due to some trigger candidates are missed, such as:

1. In the stripSegmentTool, the hash_bandid originally couldn't distinguish '1'+'72' and '17' + '2', hence the charge centroids in 2 sides only form 1 segment, causing one muon's efficiency lost with dimu samples

2. In the stripClusterTool, the judgement statement for the first time will miss the case if the first several digits with the same channelId but belonging to different bcs (the strategy allows +-1 bc now)

3. The dtheta < 15 mrad sentence caused the loop finished prematurely, resulting in the efficiency in A side get lower by 5% compared to the one of C side, fix it, and adjust its position for a more flexible performance study, without effecting the record for the later trigger logics

Before/After these changes:

./ccing @rosati @iodice @vadamico @fgravili