Reinforce misordered clusters checks

Misordered cluster checks still failed in two ways:

• sometimes clusters are duplicated by a failing front-end. For now, we ignore one of the two. This is a bit safer than doing nothing because a different fraction of the two clusters could lead to misordered x positions of clusters;
• the lastCluster (changed in this MR to lastChannel) was not refreshed when reading a large cluster. This lead to a very rare occurence where a large cluster would 'encompass' a following cluster.

Together, those changes solve the failing assert in PrStoreSciFiHits that wsa observed by @gunther on /calib/online/tmpHlt1Dumps/LHCb/0000280417/Run_0000280417_20231024-063714-898_VCEB08_0001.mdf.