Fix floating point drift in linear scale ticks

[aoright]

Fixes #12281

Motivation and Root Cause

When generating linear scale ticks with bounds === 'ticks' on non-round bounds (e.g., min: 49.894, max: 51.5264, spacing: 0.2), niceMin = Math.floor(rmin / spacing) * spacing evaluates to 49.800000000000004 due to IEEE-754 floating-point multiplication.
Because _decimalPlaces(niceMin) counts the decimal digits needed to write that value, it evaluates to 16 instead of 1.
Consequently, decimalPlaces = Math.max(_decimalPlaces(spacing), _decimalPlaces(niceMin)) becomes 16, resulting in factor = 1e16.
The subsequent tick rounding safeguard Math.round((niceMin + j * spacing) * factor) / factor then rounds to 16 decimal places, preserving floating-point drift in intermediate ticks (e.g., producing 50.00000000000001 instead of 50).
This causes custom ticks.callback functions checking integer values (e.g., value => Number.isInteger(value) ? value : '') to silently fail to match integer ticks and discard axis labels.

Changes Summary

• In src/scales/scale.linearbase.js: Normalized niceMin and niceMax to _decimalPlaces(spacing) using spacingFactor = Math.pow(10, _decimalPlaces(spacing) || 0) immediately after computing floor and ceil multiples in the bounds === 'ticks' branch. Since niceMin and niceMax are integer multiples of spacing by definition, their mathematical precision cannot exceed _decimalPlaces(spacing).
• In test/specs/scale.linear.tests.js: Added regression unit tests for positive non-round bounds (min: 49.894, max: 51.5264, stepSize: 0.2) and negative non-round bounds (min: -51.5264, max: -49.894, stepSize: 0.2) to verify that generated tick values are clean and integer ticks match Number.isInteger().

Verification Evidence

• Native typecheck passed: pnpm lint-types completed with 0 errors.
• Native linter passed: pnpm lint-js completed with 0 errors.
• Native Karma tests in Chrome passed: ./node_modules/.bin/karma start ./karma.conf.cjs --single-run --browsers chrome --grep scale.linear executed 75 of 75 tests with SUCCESS.