fix(year-level): prefer units-passed standing over calendar-year history count

The grade-history method counted distinct academic years, which over-advanced
standing for students who repeated/failed a year (e.g. showing Year 5 when the
student is genuinely Year 4, since a partial 2025 year didn't advance them).

Reorder the hybrid so the UNITS math (total prospectus units minus still-lacking
units = units actually passed) is primary, and View Grades history is only a
fallback when units data is unavailable.

parsers.js

@@ -101,12 +101,13 @@
     return history;
   }
 
-  // Standing for the study load printed at (semester, year): it is the number
-  // of distinct academic years strictly before that term's year + 1. SUMMER of
-  // year Y is the same standing as the rest of year Y (the student has just
-  // finished the preceding level during Y's regular semesters). Clamped to at
-  // most the number of distinct years in history (a student cannot be past the
-  // final level recorded).
+  // Standing for a study load printed in `year` from the View Grades history.
+  // It is the number of distinct academic years strictly before that year + 1
+  // (FIRST/SECOND/SUMMER of a year share the same standing), clamped to at most
+  // the number of distinct years in history. NOTE: this is only an
+  // approximation for the printed semester — it counts calendar years present,
+  // not units passed — so a repeated/failed year can over-advance. It is used
+  // only as a fallback when the units-based standing is unavailable.
   function yearLevelForSemester(history, year) {
     if (!history || !history.length || !Number.isFinite(year)) return null;
     const distinctYears = [...new Set(history.map((t) => t.year))].sort((a, b) => a - b);