Refactor LZ-String implementation into modular core with platform abstraction (Qt + non-Qt support)

[vakarelov]

LZString Refactor Summary

This change restructures the implementation from a single tightly coupled block into a layered, extensible architecture while preserving the public API:

Public API:
LZString::{compress, compressToUTF16, compressToBase64, decompress, decompressFromUTF16, decompressFromBase64}

---

Highlights

Core / Algorithm Separation

• Introduced LZCore to encapsulate compression and decompression logic.
• LZString is now a thin façade delegating to LZCore, improving readability and testability.
• Unified internal logic across normal, UTF-16, and Base64 variants by parameterizing bit width and output mapping.

Platform Abstraction

• Added ILZPlatform interface to abstract string length, slicing, concatenation, character code translation, and output symbol mapping.
• Implemented QtLZPlatform for QString-based builds.
• Added optional StdLZPlatform (activated via LZSTRING_NO_QT) enabling use without Qt by typedefing QString to std::string.
• Base64 alphabet and reverse lookup moved into platform objects with precomputed reverse maps.

Compression Path Improvements

• Replaced ad hoc state handling with clearer local lambdas (writeBit, writeNBits).
• Reduced duplication in emitting character codes (8-bit vs 16-bit paths) by passing a platform callback.
• Preallocation heuristics retained (reserve based on input size) to minimize reallocations.
• Dictionary and creation tracking now use std containers with explicit lifecycle.

Decompression Path Improvements

• Rewrote bit reading using compact helpers (readBit, readNBits) for clarity.
• Delegated source symbol acquisition to lambdas instead of multiple helper classes.
• Maintains original dictionary growth semantics (enlargeIn, numBits) and special handling for dictSize edge cases.

Behavior & Compatibility

• Public API unchanged; return values and edge case handling (empty input → empty output) preserved.
• Base64 padding logic retained (appends '=', '==', or '===').
• UTF-16 compressed form still appends a trailing space for compatibility.
• Numeric bounds (reset values, bit widths) match established LZ-String logic; conditional adjustments for 8-bit vs 16-bit environments included.

Extensibility & Portability

• Clear path for adding new output encodings or URI-safe Base64 variant.
• Non-Qt builds now feasible without altering algorithm code.
• Encapsulation enables future streaming or incremental APIs.

Maintainability

• Concentrated algorithmic complexity inside LZCore with a smaller, stable external surface.
• Removed repeated bit-flush sequences via reusable abstractions.
• Improved naming and structure around dictionary operations and termination markers.

Performance Considerations

• Fewer temporary string fragment objects (no substring view objects).
• Precomputed Base64 reverse map avoids per-call reconstruction.
• Structure now amenable to future switch from std::map to unordered_map if profiling justifies it.

---

Follow-up Opportunities (Not Included Here)

• Add benchmark + regression tests comparing outputs against canonical vectors.
• Introduce URI-safe Base64 variant.
• Provide streaming compression/decompression APIs for large inputs.