Bump version to 0.3.0

[mettta]

Release v0.3.0 — Pagination and Rendering Update

Summary

This release introduces major improvements across the entire HTML2PDF4Doc engine: unified slicing for Table/Grid/PRE, updated page-start logic, reworked preview/print layers, improved media handling, and expanded test coverage.

---

New Features

Unified Splitting & Slicing Engine

• Unified slicing pipeline for Table / Grid / PRE / TableLike.
• “Original-as-first” strategy for Table and PRE slicing.
• Per-slice capacity to make fit/scaling predictable.
• Normalised slice edges and cut-line adjustments.
• TD-level measurement and TD-only overflow scaling.
• Balanced row slicing for Table and Grid.
• LoopGuard to stop non-progress iterations.
• Shared recorder for slice metadata (“parts recorder”).

Page-Start & Page-Break Logic

• Updated findBetterPageStart() logic with semantic fallbacks.
• Skipping hidden, thin, or service wrappers during candidate search.
• Separate handling of first-child and last-child parent context.
• Unified measurement via getTopForPageStartCandidate().
• Correct first-page element placement.
• Reset margins on both sides of page breaks.
• Improved handling of nested tail chains and long single-child structures.

Preview & Print Layers

• Introduced the three-layer model: contentFlow, paperFlow, overlayFlow.
• Paper background moved to a pseudo-element.
• Header/footer moved to the overlay layer with transparent insets.
• Header/footer made opaque to prevent border bleed-through.
• Disabled mask in print mode (Blink duplicated text issue).
• Added bottom-edge mask for clean page boundaries.
• Frontpage inserted into content flow.
• Root element stays hidden until preview is fully generated.

Inline / Flow / Children Handling

• Standalone inline wrappers promoted to ComplexTextBlock.
• Dropping non-flow nodes early (absolute/fixed/display:none).
• Added getFirstChildrenChain / getLastChildrenChain.
• Improved thin-wrapper unwrapping via resolveFlowElement.
• Tail detection stops when encountering a text node.
• Updated getBottomWithMargin and related vertical rhythm logic.

Media

• Correct subtraction of baseline gap for inline images.
• DOM parent fallback for baseline correction.
• Improved handling of replaced elements.
• Improved tall image and SVG overflow behaviour.

Grid Engine

• Full Stage-5 splitting support, including overflow branches.
• GridAdapter: virtual rows, style caches, line-height caches, shell-height caches.
• Monotonic grid split behaviour across justify/align variants.
• Static container support and offset-based row detection.
• Width locking before splitting.

Table Engine

• Conservative ROWSPAN fallback.
• TFOOT ignored in consistency checks.
• Strict page-start registration for first slices.
• Balanced row slicing with shell-height accounting.
• Ultra-small final slices absorbed into the previous slice.
• TD-level measurement for all split decisions.
• Two-pass splitPoints with sanitisation.
• Scaling fallback for problematic TDs.

Config & Debug

• Forced Debug Mode (data-forced-debug-mode).
• Selenium-friendly forcedModeLog.
• Normalisation of string boolean config values.
• Print-mode-only test styles.
• Structured debug logging.

Testing & Tooling

• Embedded HTML page source in pytest failures for e2e tests.
• @focus mode and RUN_ALL override.
• Extended coverage: PRE, Table, Grid, ROWSPAN, IMG, nested tails, page-start heuristics, monotonic grids.
• Updated helper functions (assert_element_on_the_page, element_order, direct-children asserts).

---

Notable Fixes

• Correct baseline-gap handling for inline images.
• Fixed false empty-first-slice cases in PRE and Table.
• Correct parent-context propagation for boundary children.
• More accurate available-space estimation for page-start.
• Stable overflow fallback paths for Grid/Table.
• Correct caption/thead height calculation.
• Fixed incorrect page-start candidates caused by hidden wrappers.

---

Breaking Changes

• Stricter page-start selection may change page boundaries in some documents.