We need this higher so that we can override libraries like VTE from
application CSS.

Fixes #52

Fixes #51

suggestion: Add secondary icon support

Closes #50

See merge request !42

Fixes: #50

dzl-suggestion-row: Move from box to grid layout

See merge request !40

dzl-column-layout: Improve columns estimation

Closes gnome-builder#1037

See merge request !41

Improve the n_columns calculation in dzl_column_layout_layout() to avoid
overshooting the required width by one count of column_spacing.

Fixes: gnome-builder#1037

When we are transitioning from visible to non-visible, we have a number
of operations in play (animating, then hiding, etc). We need to do more
than just check for mapped before we short-circuit.

Fixes #49

This reverts commit 4fa4a08f.

This isn't need in it's current form for epiphany, so we'll defer until
later to see if we want to follow through on the design.

Analog to common mobile systems select whole text on first press into the
entry. Every following press will just set the text marker position.

This allows the icon to stay aligned to the first text row when using
compact-mode.

Fixes #46

We don't have any docs for this currently anyway.

Fixes #45

The stop-search signal was not being used by DzlPreferencesView and
DzlShortcutThemeEditor, and their internal GtkSearchEntry can't be
accessed by users of those widgets.

See: gnome-builder#303
Closes: #44

This is a helper to remove accels based on previously added commands. It
is mostly just useful for auto-generated keybindings situations.

This is useful so that if there are no visible widgets in the sub-page,
we don't show another subpage instead.

doc: fix typo

See merge request !36

This helps us ensure that we are matching later in the string so that
we can potentially gain a higher score (due to penalties in distance
between match characters).

We might find a word that is a better match by starting with a character
later on in the word. However, if the previous scan moved us past the
character/pos tuple that would be necessary to match, we won't find it.

This allows us to rollback the position to a location where we can once
again match upcoming character for the current key-id. It's a small hit
(walking backwards) to greatly increase the likelyhood we match a
contiguous string.

This tweaks the scoring a bit so that we can detect a perfect match (all
characters one after another) without perturbing the score with the
length. Such is necessary so that longer words with a perfect match
do not score lower than shorter works with a partial match. However, the
in-between states there are going to fall into the previous trap (which
is probably fine for now).

Fixes #41