find_exposures: unreadable HISTORY fails by accident; greedy extension strip; EXP_PREFIX = p is a no-op #908

Found while writing the epochs-from-tile-history contract for the decision record (#875), in FindExposures.get_exposure_list: (1) an unreadable tile HISTORY logs "continuing..." and then raises UnboundLocalError on hist — the tile fails, but not deliberately; (2) the extension strip is greedy, so x.fits.fz → x.fits; (3) EXP_PREFIX = p in config_tile_Fe.ini does nothing, because removeprefix never matches CFIS's trailing p (the bare number is derived later by build_index.py and make_post_process); (4) make_post_process's docstring describes ccd_size as [nx, ny] — it takes four bounds. Ordinary CFIS output is unaffected by fixing these.