Test line splitting with many runs.

Add ignored prefixes and suffixes to test data.

This ensures that implementation does not rely on the texts having zero
offset inside their byte arrays.

Remove I8/Int distinction.

The distinction was meant to provide additional type safety, but
inconsistency with other used libraries required too many workarounds
for this to be beneficial.

Define SeparableTextContainer as its own class.

Minimise working with Data.Text.Internal.

Use ICU to find line break boundaries.

Implement line breaking on code point boundaries.

Use top edge for placing lines.

Implement run splitting.

Lay out Runs independently of Spans.

Store span index in ResolvedSpan.

Link to Span using a generic wrapper.

Calculate fragment position continuously.

This merges the various "arrange" functions into the layout functions,
which will be used to detect overflows for line breaking.

Implement absolute line heights with half-leadings.

Calculate "normal" line height and pen position.

Use CSS-compatible line height settings.

Fix pretty printing for fragmented spans.

Upgrade Fragment to contain pen position.

Use CSS3 terminology for fragments and containing block.

Avoid reimplementing Data.Text.Lazy.fromStrict.