Move most of maintenance mode logic into maintenance_mode module

[palant]

Description

This establishes the same internal API for maintenance mode as for health and metrics. While at it, I adjusted get_response() to make producing body_content more straightforward – no long a mutable variable that might or might not be overwritten but two straightforward branches. Side-effect: empty maintenance mode file will no longer result in default message. IMHO this makes sense because a) someone might actually show an empty message and b) the message about falling back to default message is only logged when the maintenance mode file doesn’t exist, not when it is empty.

I also changed DEFAULT_BODY_CONTENT because the old message is IMHO grammatically incorrect. It should be either “server is under maintenance” or “server is in maintenance mode” (I went with the latter).

Related Issue

This is another step towards fixing #342.

How Has This Been Tested?

I added some rudimentary unit tests. Functional testing on Linux was also successful.

[semanticdiff-com]

Review changes with SemanticDiff.

Analyzed 3 of 3 files.

Overall, the semantic diff is 9% smaller than the GitHub diff.

	Filename	Status
✔️	src/handler.rs	13.84% smaller
✔️	src/maintenance_mode.rs	5.75% smaller
✔️	src/server.rs	20.13% smaller

[joseluisq]

Thanks!

[joseluisq]

Side-effect: empty maintenance mode file will no longer result in default message. IMHO this makes sense because a) someone might actually show an empty message and b) the message about falling back to default message is only logged when the maintenance mode file doesn’t exist, not when it is empty.

Fair point.