v1.1.0: refactored cams parser

New

• CASData.parse_warnings. The CAMS/KFin DETAILED parser now reconciles each
  scheme's transactions against its printed running Unit Balance column — the
  statement's own checksum — and records a non-fatal warning on any
  discontinuity (prev_balance + units != printed balance) or closing-balance
  mismatch. A non-empty list means a transaction row was likely dropped or
  mis-parsed; the parse still returns, but that scheme's data should not be
  trusted blindly. The region-based header parser additionally warns when a
  scheme-header region is discarded without its Opening Unit Balance anchor
  (folio/AMC boundary, end of document, or an unparseable header) — previously
  such a scheme and all its transactions vanished silently. Warnings are printed
  by the CLI after a successful parse.
• JSON Schemas for the output models. schema/CASData.schema.json and
  schema/NSDLCASData.schema.json are generated from the Pydantic models
  (scripts/generate_schema.py); the README shape documentation was reworked
  around them.

Fixed

• Dividend reinvest misclassification. DIVIDEND_RE carried an inline
  (reinvest)* group inside two lazy quantifiers that never backtracked to
  capture, so "Reinvestment of IDCW @ Rs…" and "IDCW - Reinvest @ Rs…" were
  classified as DIVIDEND_PAYOUT. Reinvest is now detected with a separate
  substring search.
• Folio collisions across AMCs. Folios were keyed by folio number alone, but
  folio numbers are RTA-scoped, not globally unique — two AMCs sharing one
  silently merged the second AMC's schemes into the first's folio. Now keyed by
  (amc, folio_no).
• Advisor-wrap header mis-parsed as a bogus ARN scheme. A wrapped
  (Advisor: ARN-xxxxx) continuation line could anchor a spurious scheme whose
  RTA code parsed as ARN, splitting one holding into an empty stub plus a
  mis-coded duplicate.

Changed

• Scheme-header parsing rewritten as anchor-delimited regions. The CAMS/KFin
  DETAILED parser no longer stitches headers by line-local proximity
  (Y_BAND / consumed_below lookahead). It accumulates the lines between the
  folio line (or the previous scheme's footer) and the next Opening Unit Balance into one region and parses fields from it once — bounded by the
  document's reliable single-line anchors, so it tolerates arbitrary header wrap
  distance and stitches headers across page breaks (which the old code could
  not). Output is compatible with 1.0.1; the only changes on the private corpus
  are cleaner scheme names (interleaved-ISIN IDCW names keep their
  Payout / Reinvest qualifier; leaked advisor / ISIN fragments are removed).

Full Changelog: v1.0.1...v1.1.0