b% dt set late (or not at all?) in mesa/binary

**Describe the bug**
In preparing for this year's MESA summer school massive star binary lab, I'm noticing a mismatch between the dt reported by each star (b% s1% dt and b% s2% dt), and the binary itself in mesa binary (b% dt). Specifically, while the s% dt 's are nonzero and commensurate with the timesteps reported in the terminal, b% dt remains zero.

I don't think this will affect the evolution, obviously, as each star is taking a timestep, but it does affect run_star_extras routines which attempt to modify or utilize the timestep in any calculations.

**To Reproduce**
Example steps to reproduce the behavior:
1. in $MESA_DIR/binary/test_suite/evolve_both_stars/src/run_binary_extras.f90 (and/or other binary test suite directories), edit extras_finish_step to include the line (e.g.)
write(*,*) 'star1:', b% s1% dt, 'star2:', b% s2% dt, 'binary:', b% dt
3. ./mk && ./rn
4. While star1 and star2 have nonzero timesteps, b% dt is = 0...

**Expected behavior**
I would expect b% dt to be nonzero and equal to the two stars' timesteps, and I would expect that modifying b% dt in run_binary_extras.f90 's extras_finish_step would affect the next timestep. At this point it seems decoupled from the evolution.

**Screenshots**

**Desktop (please complete the following information):**
- OS: mac OS Ventura 13.4; Apple M1 Max chip
- mesa-r23.05.1 with the most recent sdk