fix #41761: beam angle too sharp

[MarcSabatella]

After reading a number of recommendations, I opted for a fairly subtle fix, just limiting the beam slant to 1sp for two-note beams. Gould talks of limiting it further depending on how crowded the spacing is, but that's not easy for us to do.

BTW, while fixing this, I discovered a couple of places where our beams were really strange. The first measure of the vtest I created actually shows cases where, in addition to issues with the beam angle, the stems were too short or two long overall. I think there were just some bad values in the table of beam metrics. I didn't do a super through job of it, but I tried different intervals transposed to different staff lines, and I fixed the most obvious cases.

This is not an area of code I know well, so I definitely would appreciate review, but it does seem to do what it is supposed to do. I tried changing the sixteenth note portion of the code to limit the slant to 0.5sp, but the subsequent code that calculates stem lengths doesn't seem to handle this, so I arrive at the values for minS and maxS I have empirically. I'm less confident about that paet of the change than the eighth note part..

[MarcSabatella]

So it turns out the beam mtests are doing their job well - testing lots of combinations of beams and verifying the metrics. I checked them all out by eye and made a few more improvements (I hope) to the table, then updated the tests. So loading the beam tests A-G (which were the ones affected) would be a good way for anyone to check for themselves the different combinations.

[lasconic]

I cannot reproduce the tests failure if I merge the pr locally, so I wouldn't worry for now.

[MarcSabatella]

It works for me locally too. I've been working on the regression from my other change and haven't looked into this anyhow.