Update scheduler, global and parallel updates, measurement fixes/changes

[ffreyer]

Global updates

• adjust a bunch of methods to work with a given configuration rather than just mc.conf
• add reverse_build_stack because the best point to do a global update is at current_slice = 1; direction = 1 (because greens has just been built at this point)
• add a torn-apart calculate_greens that yields a diagonal matrix usable for the determinant ratio calculation
• add propose_global_from_conf and accept_global
  • det(G) is less stable than prod(mc.s.Dr) which appears during the calculation of G. Note that using just the diagonal matrices means ignoring a potential complex phase from unitary matrices. This phase factor should be 1 unless we have a sign problem, which we already check for in local updates. So I think it's fine to ignore.
  • It may be useful to have deny_global instead deny_global is more work than accept_global because both require a full stack rebuilt (either new or old) but deny also requires finishing the greens calculation.
• add global_move
• add update scheduler
  • the (original) point of these is to (1) be able to mix different global updates and (2) to dynamically phase out bad global updates. The AdaptiveScheduler can do both.
• add various update structs
  • GlobalFlip suggests conf -> -conf
  • GlobalShuffle suggest conf -> shuffle(conf)
  • NoUpdate suggests doing nothing (needed as counterbalance for adaptive scheduler)
• move global stats to updates update wrapper AcceptanceStatistics
• remove global_rate
• ReplicaExchange and ReplicaPull updates
• split global.jl into multiple files - i.e. scheduler.jl, global_updates.jl, parallel_updates.jl (and eventually local_updates.jl)
• compress scheduler prints
• let the scheduler take care of all (i.e. also local) updates?
  • use LocalSweep() and LocalSweep(N) rather than LocalUpdate/LocalSpinFlip
  • generalize global_update to update, move temps to DQMC/stack
• move temp conf to DQMC

Measurement changes

• add ApplySymmetries to process symmetry calculations during the simulation
• move greens and lattice iterator types to fields of DQMCMeasurement

General changes

• split up DQMC.jl into a bunch of other files. DQMC.jl is now ~500 lines instead of 1000, the other files are typically 100-200 lines. Bit cleaner imo. Also cleanup comments/docstrings a little
• fix parameters (linear system size -> total sites)
• simplify DQMC construction (by making all concrete types known beforehand. There are some related changes here and there)
• some QoL changes (DQMCParameters reconstruction, easier application of Sum wrappers)
• update tests to Julia 1.5/1.6
• Access d.diag instead d in vmul! methods for d::Diagonal
• update tests

[ffreyer]

+1000 -150

Oh god it's happening again

[ffreyer]

Tests passed locally, Should pass here after https://github.com/JuliaRegistries/General/pull/33584 is done

[ffreyer]

closes #105