CSV/RTF: bound cell/text materialization to prevent memory exhaustion

## Problem
The CSV and RTF parsers appear to materialize cells/text without a bound on the number of cells or the decompressed text size. A crafted (or accidentally pathological) CSV/RTF input could exhaust memory during conversion.

## Expected outcome
Enforce limits analogous to the existing `max_entry_bytes`/resource-limit checks (or a row/cell count cap) for CSV/RTF, surfacing a `resourceLimit` error instead of unbounded allocation.

## Context
Raised while adopting `firecrawl-anydoc` 0.1.9 as a Python dependency (GroktoCrawl parse-svc). Applies to the Rust core and all bindings. A 50 MB request cap in the consuming service is only a partial mitigation.

Start by locating the CSV and RTF parsers and the existing max_entry_bytes/resource-limit checks in the Rust core. Exercise the parsers with oversized or high-cell-count inputs, then verify that bounded processing returns a resourceLimit error across the Python and Node.js bindings.

Written by the indexing model from the issue text.