Add pyjhora_batch: headless batch horoscope reports (PDF/TXT/JHD)

[jivrajss]

Summary

Adds pyjhora_batch, a headless batch layer that turns a CSV/Excel of birth
details into per-person horoscope reports. It reuses PyJHora's engine
throughout and reimplements no calculations.

The report is extracted once as structured data (report_data.build_report)
and then rendered, so no output path needs Qt on screen:

Output	Written by	Notes
.txt	text_writer	wrapped plain-text report
.pdf	pdf_writer	vector PDF via reportlab — ~150 KB, searchable, selectable
.jhd	jhd_writer	Jagannatha Hora seed file, importable into the app

Notable points

Vector PDF instead of a widget screenshot. The original path captured the
Qt widget at ~130 DPI and embedded it as JPEG — roughly 10 MB per person,
blurry, not searchable. Typesetting the same data as real text gives ~150 KB.
The screenshot path remains as --pdf-mode screenshot, since it is the only
one that includes the drawn chart diagrams.

.jhd encoding. Every angular/temporal field packs as sexagesimal
D.MMSS (the digits after the point are literally minutes and seconds),
longitude and timezone are East-negative, latitude is North-positive, and
lines are CRLF. The person's name is never written to the file — JHora keeps it
in the filename. Derived byte-for-byte from a real export read back against the
app's own rendering of it.

Yoga tables report the rule, not a verdict. yoga_msgs_<lang>.json carries
a third field per yoga alongside name and condition. It is fixed boilerplate
keyed on the yoga name — identical for every chart the yoga fires in — and
several yogas match very loosely. _matrunasa_yoga_198_calculation returns
true whenever the Moon is hemmed by, associated with, or aspected by any
natural malefic, which covers a large majority of charts;
mathibhramana_yoga_294 ends in "aspected by or conjoined with any other
planet". Rendered per person next to a name and birth time, "The person's
mother will have a very early death" and "The person becomes insane" read as an
individual prognosis rather than as the classical rule text they are. The
tables show the yoga and the condition that matched, and note that only present
yogas are listed.

Happy to make that last one opt-in behind a flag instead if you'd rather keep
the current output as the default.

CLI

python -m pyjhora_batch births.csv -o reports

--no-pdf / --no-jhd / --no-txt to skip outputs, --pdf-mode {vector,screenshot}, repeatable --dhasa NAME (default vimsottari;
--dhasa all for every system PyJHora exposes), -w N for parallel workers.
Per-record failures are isolated and summarised rather than aborting the batch.

Dependencies

reportlab>=4.0 for the vector PDF. openpyxl (Excel input), pytest and
pypdf (test suite) stay optional — .txt and .jhd need nothing beyond
PyJHora itself.

Testing

169 passed, 1 skipped. Fixtures use a synthetic reference chart. The panchanga
and D1 expectations are captured from ChartTabbed.compute_horoscope rather
than from this code's own output, so they cross-check the headless path against
the GUI — in particular they catch a skipped drik.set_ayanamsa_mode, which
otherwise shifts every longitude silently.