Add AffineConstraint::add_constraint().

[bangerth]

For a long time, I had wished that AffineConstraint had a function to add a whole constraint all at once, rather than doing it via add_line(), add_entry(), set_inhomogeneity(). This has also been mentioned in #334, and it has come up in #15375. This patch implements it.

Rather than write a new test, I decided to just convert a whole bunch of places in functions we use extensively. Also a few of the tutorials.

In the end, I would love to deprecate the three functions mentioned above. I think that interface is just not very good. What do people think about that?

[gassmoeller]

I really like this addition. When I learned about the individual functions I found it quite unintuitive that they were not combined in one call. I find the combined function way easier to understand. One question about the documentation.

[gassmoeller]

typically we list contraints as u_i ..., here you capitalize U. Was this on purpose?

[bangerth]

I switched it to x_i since that's what the documentation of the AffineConstraints class uses.

[kronbichler]

In light of #16121, I believe this is wrong and should be add_entries(...). I will make a quick test to see if it works, otherwise I have to postpone it because my time is constrained.

[bangerth]

I have a follow-up patch that avoids the repeated calls. Hold on with your timings.

[kronbichler]

In the end, I would love to deprecate the three functions mentioned above. I think that interface is just not very good. What do people think about that?

I think it would be good to deprecate those functions eventually indeed, or at least limit them to internal use. So this has my support. We will need some time to let users adapt to this change, because it will affect many users.