Fixup of #18270

Summary of the issue:
We use gettext in l10nUtil now, but it wasn't bundled correctly into
NVDA

Description of user facing changes:
l10nutil is fixed for translators

Description of development approach:
bundle gettext into nvda distributions

Testing strategy:
test with nvda build

Known issues with pull request:
none

Part of #17878
Follow up to #18270
Summary of the issue:

Currently, fetching the latest translations from Crowdin is a manual process.
Description of developer facing changes:

    A GitHub action to update translations is added. This will run at weekly interval, just before we do a regular release. Translation freezes should be designed to line up with this time. This can be run manually. This will only update tracked translations for existing languages.
    A GitHub action to add new languages is added. This is because we have languages which are started in Crowdin, but not ready for sync/release. This will allow us to track new languages.

Description of development approach:

    The Crowdin GitHub actions caused problems when running in a windows environment. Using the API directly with l10nUtil seemed cleaner.
    To download all translations you must create a project export bundle and download it directly. Adding this to l10nUtil makes it easier for devs to pull all translations from crowdin manually when required.

Fixup of #18503 and #18270
Summary of the issue:

The regex to check for unnamed string interpolations doesn't support the %% symbol, which escapes %.
This leads to certain usages of it raising incorrect interpolation errors.

File source/locale/tr/LC_MESSAGES/nvda.po: 1 error
Message starting on line 10286
Original: "%s%%"
Translated: "%%%s"
Error: unnamed percent interpolations differ
Expected: unnamed percent interpolations in this order: ['%s']

The regex for missing names for brace interpolations is not working.
i.e. "{}" should raise an error without a name like "{name}".
This causes issues where the order of the arguments change in the string.
e.g. "Character: {}\nReplacement: {}" being translated to "Replacement: {}\nCharacter: {}" will result in the expected interpolation being in the wrong place.
These should be fixed in the source .po files to add names to instances of "{}".
Description of user facing changes:

Adds a more rigourous check for interpolations using brace formats, which will reduce string errors in the UI.