fix(text): reset continued state when an explicit position is given (#1641)

[JamBalaya56562]

Summary

Fixes #1641.

When a text() call passes an explicit x/y coordinate and is not itself continued, it now discards any wrapping state left behind by a previous continued: true run and draws at the absolute coordinate you pass — instead of being offset by the previous run's inline position.

The problem

continued is a forward-looking flag: the state is saved based on the current call's flag and reused by the next call (lib/mixins/text.js#L91-L92), and the next _text reuses this._wrapper unconditionally at L84. On continuation the saved continuedX is added to the current x in the wrapper's firstLine handler (line_wrapper.js#L48-L49, accumulated at L352). So a coordinate passed to the call immediately after a continued run is offset, not absolute:

for (const t of ['AAA ', 'BBB ', 'CCC ']) doc.text(t, { continued: true });
doc.text('FRESH', 400);   // requested x = 400

	x of FRESH
Before	484.02 (= 400 + accumulated inline width)
After	400

Previously the only way to avoid this was to set continued: false on the last segment of the run, which is awkward inside loops and when delegating text to helper functions.

The fix

A guard at the top of _text: if a positional x or y is given and the call isn't itself continued, clear this._wrapper / this._textOptions before initializing options. continuedX lives on the wrapper instance, so dropping the wrapper resets it to 0 naturally.

Why this is safe (not a breaking change)

The canonical rich-text idiom continues to work unchanged, because continuing segments pass no coordinates:

doc.fillColor('green').text(a, { width: 465, continued: true })
   .fillColor('red').text(b);   // no x/y -> still continues inline

Verified against the rest of the codebase:

• Existing continued unit/visual tests and the docs/text.md examples pass no coordinates on the continuing call → unaffected.
• heightOfString already nulls this state today → behavior identical.
• boundsOfString uses a local wrapper and never touches this._wrapper → unaffected.
• list uses its own wrapper via _initOptions → unaffected.
• table renders each cell at explicit coordinates and cells don't continue → unaffected (and now protected from state bleeding into a cell).

Tests & docs

• Added two regression tests in tests/unit/text.spec.js: explicit position resets after a continued run; a continued run without coordinates still flows inline.
• Documented in docs/text.md that continued is forward-looking, how to end a run, and that an explicit position starts a call fresh.
• CHANGELOG entry under Unreleased.

Full tests/unit + tests/visual suites pass (the one pdfmake/tables failure is a pre-existing default-5s-timeout flake that also fails on master and passes with a longer timeout — unrelated to this change).

🤖 Generated with Claude Code