Add basic config mechanism for fido2

[robin-nitrokey]

This patch adds a basic config mechanism using the admin app. The configuration is loaded and applied during initialization. The admin app now has a special role, is constructed first and can no longer be disabled. Other applications can receive a reference to the configuration loaded by the admin app.

To avoid overwriting configuration values for apps that are not enabled in the current firmware, the components of the Config struct are not feature-gated. It could be simplified by using a derive macro but that seems overcomplicated for a first implementation.

The first use case for the config mechanism is enabling or disabling skipping the FIDO2 user presence check directly after boot.

To do

• Add test for serialized config size

[robin-nitrokey]

One open question is how to deal with default values and what should happen if we decide to change them in future firmware versions. We could skip serializing the application config if it is equal to the default. This would mean that a firmware update with changed default settings would change the existing configuration.

[daringer]

Generally we should not alter configuration values if they've been set by the user. So skipping serialization if it's the default is ok, once the user changes it, it would automatically persist, right?

[robin-nitrokey]

I was thinking about cases where the user deliberately sets the default config. We could add an additional field to indicate manually written configuration, but that would increase the serialized size.

[daringer]

Ultimately, this is a (client) UI issue. Means I would vote to keep the firmware implementation as simple as possible (as you suggested). Our client software (pynitrokey / nitrokey-app2) should on top then make sure that the user is always well informed. For nkapp2 this means the configuration has to be shown after updating and pynitrokey should do something similar after updating.

[robin-nitrokey]

Changes since the initial version:

• Don’t write default values.
• Shorten field names to reduce config size.
• Test that the maximum configuration is smaller than 1/4 IFS block size.
• Change enable_skip_up_timeout to disable_skip_up_timeout: enable would make more sense, but using disable allows us to have false as the default value, i. e. this field can be skipped during serialization.

[sosthene-nitrokey]

LGTM

[sosthene-nitrokey]

Can we merge this? That way I can rebase #335 on top of it.

Since the admin-app changes have already been merged this makes it easier to update #335