The main difference between g_hash_table_insert() and
g_hash_table_replace() is which key will get freed when replacing a key.

We want the key to also be replaced since the data owns the key.

We have the expectation the new key should be stored.

If the primary module has an env associated with it, apply it when
preparing for build so that it gets picked up.

`path()` is deprecated

plugins/meson-templates: Use `full_path()`

See merge request GNOME/gnome-builder!783

Appstream decided to use reversed domains. Please merge before 46 if possible.

metainfo: Fix wrong developer id

See merge request GNOME/gnome-builder!784

That way we don't have broken ctrl+p.

Fixes #2169

Added shortcuts for submitting the Create New Project form by Ctrl+Enter
anywhere in the form or activating the text fields with Enter/Return.

Closes #1906

Initiative to update app metadata URLs, see https://gitlab.gnome.org/GNOME/Initiatives/-/issues/51 for background

create-project: Add shortcuts to submit the Create New Project form

Closes #1906

See merge request GNOME/gnome-builder!785

metainfo: fix URLs

See merge request GNOME/gnome-builder!786

Related #2166

We keep the object-tree alive while processing, which probably should be
addressed at some point. (It does clean up at the end). But this helps
drop a bunch of resources as soon as they are no longer necessary.

With this and the previous patch I can keep the kernel to around 1Gb
resident while building the search indexes, and have it drop back down to
where it should be afterwards.

This likely fixes the majority of #2166

Fixes URL's which were not replaced in !786

metainfo: fix rest of URLs

See merge request GNOME/gnome-builder!787

Help avoid damaging the border so we get better performance while scrolling
with the source map visible.

This has some work to tweak the sourcemap in such a way that it can
render faster in a number of scenarios. Particularly around fractional
scaling and changes near the border.

GTK 4.14.1 is going to have considerable performance improvements we've
been working through. This shouldn't warrant the same necessity that it
did previously.