prohibit use of regression_l1 objective with linear trees

[jameslamb]

From the very first PR where @btrotta introduced linear trees here (#3299), it was never intended that they'd be used with the L1 loss. However, that's been technically possible the entire time because of a typo here:

LightGBM/src/io/config.cpp

Lines 431 to 433 in 6437645

	if (objective == std::string("regresson_l1")) {
	Log::Fatal("Cannot use regression_l1 objective when fitting linear trees.");
	}

This fixes that typo, and adds a test ensuring that LightGBM raises the expected errors when using unsupported parameters with linear trees.

Notes for Reviewers

This bug was caught by the newest version of crates-ci/typos... pretty cool to see a spellchecker catching a real bug!!! (cc @borchero )

This PR also updates other pre-commit hooks to their latest versions (though they didn't catch any new issues).

[StrikerRUS]

Very nice, thank you!