nice, will test on mainnet.

unbreak unit test

add unit tests for format-strings linter

use optional<string> insteald of string *

The more immediate reason I'm backporting this is that the new flag is used for the new -loglevel= arg added in https://github.com/bitcoin/bitcoin/pull/25614

tldr; the current typing system is basically just descriptive but not enforced. The only check that is being done is that INT args are not negated.
It would be good to be able to switch to stronger typing in args, but doing so would likely break many user settings, so it will have to be done carefully on existing args. This change reflects the current state of type checking: most args can be of any type.

fix test by not dereferencing std::nullopt, use .value() in that test instead of the deref operator so that we get an error is value is unexpectedly std::nullopt

There is something weird happening with clang. The test failures show that sometimes the value is the one from the previous test when we expect it to be the empty string. And the previous test does not reuse the same var, so somehow clang does not initialize std::optional<std::string> to std::nullopt but jsut reuses the same memory buffer and keeps the data to initialize the value var.

/work/src/test/util_tests.cpp(592): error: in "util_tests/util_ParseKeyValue": check *value == "" has failed [v != ]
/work/src/test/util_tests.cpp(606): error: in "util_tests/util_ParseKeyValue": check *value == "" has failed [1 != ]

hard to see what the error is, in the logs of the clang build. Will have to reproduce locally

remove unused function declarations. In the source material these did not exist because the functions were used as static (out of order backports, args.cpp split into multiple submodules)

fix stuff

Not sure if this helps, but one of the issue with tx determinism in the test framework comes from random data being used for padding transactions to 100 bytes.
In D17933 I added a deterministic arg to pad_tx to make it use deterministic data.

Looks good.

rebase. The error is caused by another commit (not yet submitted as a diff)