Fix timesearch near end of title
When searching by time on a call to dvdnav_jump_to_sector_by_time(), eventually we hit the function dvdnav_find_vobu_by_tmap() in libdvdnav/src/searching.c.
This locates the surrounding tmap entries (each of which can cover multiple vobus, listing the start sector of specific vobus at fixed time intervals) for the start sector of the cell containing the target time, then calls dvdnav_tmap_calc_time_for_tmap_entry() with them and the cell start sector to interpolate the start time value of that cell, which we then use to do a relative seek to the right vobu within the cell. I think...
calc_time itself looks up the vobu indices in the admap (which lists the start sector of every vobu) from the incoming tmap sector numbers, then interpolates the cell start time from those vobu indices.
The problem is that the admap search can return "fake" end-markers, presumably to ensure an out-of-bounds seek returns something sane. These appear to have the same vobu_idx as the first or last "real" entries, so it can happen that when you search for the surrounding range you end up with start and end values with the same vobu_idx. The interpolation then substracts one from the other, and divides the current offset by that, triggering a divide-by-zero FPE.
(This can happen for example when hammering the step-forwards button (usually right-cursor) repeatedly when already close to the title end.)