Rparadox 0.3.0

Major changes

Complete rewrite of the C backend: removed all write functionality, making the library read-only. Removed dead write-code including PX_create_file, PX_put_record, PX_update_record, PX_delete_record, PX_pack, and all encoding write functions.
Removed GSF (GNOME Structured File) and memprof support.
Removed paradox-gsf.h, paradox-mp.h, px_encode.c/h, px_memprof.c.

Fixed uninitialized memory access in get_px_head() that caused a segfault with -O2 + LTO when opening encrypted files.
Added missing #include "paradox.h" in 5 .c files (paradox.c, px_io.c, px_head.c, px_misc.c, px_memory.c).
Fixed buffer overflow in px_passwd_checksum() for passwords > 255 chars.
find_blob_file() now normalises paths with forward slashes to ensure consistent expect_equal() comparisons on Windows UCRT.
Fixed 238 Valgrind errors (uninitialised reads) in get_px_head() and _px_get_data_blob(): replaced ret < 0 checks with exact-size comparisons (ret != sizeof(...)) after read() calls to detect short/truncated reads from corrupted files.

README.Rmd: removed all runnable code examples, now links to vignette.
Vignette: now contains all usage examples previously in README (magick plot, encrypted files, encoding override, advanced workflow).

Fixed rchk warnings: “Suspicious call (two or more unprotected arguments) to Rf_setAttrib” in pxlib_get_data_c() (interface.c:341, 347)

Fixed rchk warnings:
- Suspicious call (two or more unprotected arguments) to Rf_setAttrib at pxlib_get_data_c Rparadox/src/interface.c:317
- Suspicious call (two or more unprotected arguments) to Rf_setAttrib at pxlib_get_data_c Rparadox/src/interface.c:326
Optimized performance using local static variables for constant caching
- Reduction in allocVector() calls for repeated file reads
- Caches class vectors: c(“pxdoc_t”, “externalptr”), c(“hms”, “difftime”), c(“POSIXct”, “POSIXt”)
- Caches name vectors for metadata structure
Added download statistics badges to README (total, weekly, daily)

Encryption Support: Added the password argument to read_paradox(). Users can now read password-protected Paradox files (.db) and associated BLOB/Memo files (.MB).
Enhanced Metadata: pxlib_metadata() now returns human-readable field types (e.g., “Alpha”, “Date”, “Memo”) instead of raw integer codes.
Encoding info: Metadata now explicitly reports the detected encoding.

Major x64 compatibility update: Integrated patches for the underlying pxlib engine to ensure stability on 64-bit systems.
- Replaced long int with int32_t in data reading functions to ensure correct data width on 64-bit systems (Windows/Linux).
- Fixed pointer arithmetic logic in src/px_head.c to prevent undefined behavior and header corruption.
- Updated writeproc signature to ssize_t for POSIX compliance.

Added a high-level read_paradox() function for a simple, one-step workflow to read Paradox files. This is now the recommended function for most users.
Added a pxlib_metadata() function to retrieve database metadata (number of records, field details, encoding) without reading the entire dataset.

The architecture for handling character encodings has been completely refactored. The C-level code now returns raw character data, and all conversion to UTF-8 is handled reliably in R via the stringi package(ICU) instead of Riconv. This fixes test failures on CRAN and improves robustness for files on various operating systems.
Added stringi to package imports.
Added extensive tests for the new read_paradox() and pxlib_metadata() functions.
Documentation and examples have been updated to reflect the new functions and recommended workflow.

Fixed CRAN check issues on Debian: package no longer attempts to open Paradox files in write mode (rb+ → rb). This ensures full compliance with CRAN Policy (no writes outside of temporary directories).
Fixed heap-buffer-overflow in PX_get_data_alpha (discovered with AddressSanitizer). Now the function safely respects field length and prevents out-of-bounds reads for fixed-length strings without \0.
Improved handling of invalid/null Paradox dates with an upper bound filter for extreme values.
Added proper .gitignore entries for cran-comments.md and CRAN-SUBMISSION.
Updated README with CRAN badges and installation instructions.
Expanded test coverage: re-enabled datasets with German, English and Russian encodings, including CP866.

Fixed CRAN feedback:
- Removed unnecessary \dontrun{} in examples.
  Added fast, self-contained examples using demo files
  (biolife.db, of_cp866.db in inst/extdata).
- Ensured no commented-out example code remains.
- All examples now run in < 5s and without commented-out code.
Fixed vignette build issue on CRAN check servers:
- Explicitly added rmarkdown in VignetteBuilder
  to avoid failures when running with --no-suggests.
Memory management improvements:
- Fixed memory leaks in src/interface.c when reading BLOB fields.
- Examples and vignette now run cleanly under Valgrind.
Verified that all examples run in < 5 sec and vignettes build
successfully across test environments.