Collect an APAAR ID at first login, and fix three student-upload faults

[vikrantwiz02]

Frontend counterpart: FusionIIIT/Fusion-client#294.

APAAR ID at first login

apaar_id joins the first-login profile popup as a required, 12-digit field. profile_completed only flips on a successful submit and the popup blocks the app until then, so no student reaches Fusion without supplying one.

Verified end to end against a real record (a roll number), rolled back afterwards:

GET exposes apaar_id: True
no APAAR ID          -> 400  APAAR ID is required
APAAR ID too short   -> 400  APAAR ID must be exactly 12 digits
valid APAAR ID       -> 200  Profile completed successfully
saved: apaar_id='a 12-digit id'  profile_completed=True

The column is on both admission models. The popup resolves a record from either table and then serializes and saves through shared attribute names, so a UG/PG-only column would have raised AttributeError for a PhD student. It is also threaded through the admin paths so it is not write-only: the spreadsheet parser accepts an APAAR heading, both bulk creates persist it via the shared _common dict, the manual add and update_student carry it, and get_student and the batch-student list return it. The download template is deliberately unchanged — the value comes from the student, not the office.

Three tests added: required, 12-digit format, and the happy path asserting the stored value.

Three faults found while uploading this year's intake files

An extended heading stopped a column being read. The parser matches lowercased headings against exact spellings; this year's template says JEE App. No. / CCMT Roll No. / AICTE App. No. where the list has jee app. no. / ccmt roll no.. Every row of a 554-student upload therefore had an empty jee_app_no — which also silenced duplicate detection, since that check keys on the application number. A field left unmatched now falls back to a heading that contains one of its spellings (aliases under 8 characters excluded, no column claimed twice). Re-parsed: jee_app_no: 'a 12-digit id'.

A repeat upload reported a raw unique-index violation. The duplicate check read camelCase keys while a parsed payload carries the spreadsheet's headings, so it compared nothing; and it only ran when skip_duplicates was set, leaving the insert to fail on the index. Duplicates are now always detected, with skip_duplicates deciding only whether they are passed over or reported:

	before	after
default	duplicate key value violates unique constraint "…roll_number_key"	Roll Number a roll number already exists for a student
skip_duplicates: true	same Postgres error	2 duplicates skipped

An M.Tech Design intake could not be saved at all. A specialization-to-discipline table overrode the discipline the sheet states:

elif specialization == 'Design':
    discipline_obj = get_or_create_discipline('Design')

so the lookup searched for M.Tech Design under discipline Design, while that batch is run by Mechanical Engineering — and the failure message concatenated batch name and guessed discipline into Create batch for M.Tech Design Design, naming a batch that cannot exist. When the strict lookup misses, the sheet's own discipline is now preferred, then a batch of that name if it is the only one for that year, and the message names a real batch and lists the disciplines that run it. Only reached when the current lookup already failed, so no upload that works today changes behaviour.

With the real file: save -> 200 '1 students uploaded successfully.', stored branch='Mechanical Engineering' spec='Design' programme=pg year=2026. The same table also pins Mechatronics, CAD/CAM, Manufacturing and Automation, Data Science, AI & ML and the ECE specializations; they agree with the current batches, but the fallback now covers them if a department ever changes.

Checks

Every path exercised against a copy of production data with all writes rolled back; manage.py check clean, no missing migrations. Needs python manage.py migrate on deploy (0052_student_apaar_id) before the frontend ships, or the profile submit will 500 on the missing column.