Fix misc compiler warnings

[jschmidt-icinga]

Together with #9411, #9728, #9729 and #9730 this fixes most of the remaining compiler warnings, except the following:

• A few -Wdangling-reference in scheduleddowntime-apply.cpp, which I'm pretty sure are false positives.
• The OpenSSL 3.0 -Wdeprecated-declarations warnings.
• The '-Wdeprecated' warnings about %pure-parser being deprecated in the .yy parser files. At first I wanted to include them here, but despite things seemingly building fine it seems there's some additional implications with replacing %pure-parser with %define api.pure and %error-verbose with %define parse.error verbose that I can't judge because I'm not a bison expert.
• Maybe additional warnings on MSVC if more than the default are turned on (/Wall)

[Al2Klimov]

Note: This currently duplicates some of the efforts in #9411 until that is merged.

Suggestion:

1. Rebase #10609 on top of #9411
2. Change your base branch from master to compiler-warnings

[yhabteab]

• The '-Wdeprecated' warnings about %pure-parser being deprecated in the .yy parser files. At first I wanted to include them here, but despite things seemingly building fine it seems there's some additional implications with replacing %pure-parser with %define api.pure and %error-verbose with %define parse.error verbose that I can't judge because I'm not a bison expert.

Oh, forgot to send my reply to this. These warning don't exist on my machine and replacing %pure-parser with %define api.pure {,true,full} results in a bison syntax error. I use the following bison version:

$ bison --version
bison (GNU Bison) 2.3
Written by Robert Corbett and Richard Stallman
---

And after going through the Bison docs, there's quite no difference between %pure-parser and %define api.pure true. There seems to be a difference between full and true using as a value though, but I guess that's irrelevant now, since we won't change this anytime soon.

[jschmidt-icinga]

• After switching the compiler to clang it found a few new variations of unused variables which I fixed as well. Maybe we should add a runner that also tests with clang to get those warnings too.
• Hopefully the windows warnings about branches not returning is now fixed by making the return type of icinga_assert_fail() function void. (Edit: Seems MSVC is still not happy)
• I've changed the timestamps used for the replay log file names from using int to std::uint64_t. I didn't just leave them as double so they can easily be converted from and to strings from/for the file names.

[jschmidt-icinga]

And the dumbest compiler award goes to... *drumroll* 🎉MSVC🎉, for the inability to infer that the expression (false ? 0 : noreturn_function()) does in fact not return. Something something industry standard.

[yhabteab]

LFTM! Ty.