Deduplicate shared preamble in expectation-value sweeps

[yurekami]

Fixes #4209.

What

DensityMatrixSimulator.simulate_expectation_values_sweep and
Simulator.simulate_expectation_values_sweep_iter began with an identical preamble
(the one flagged by the # TODO(#4209) comment): reject terminal measurements unless
permitted, resolve the qubit order, build the qubit→index map, and wrap the observables
as cirq.PauliSums.

This factors that preamble into a protected helper _qubit_map_and_pauli_sums on
SimulatesExpectationValues (the base both simulators already inherit) and removes the
# TODO(#4209) marker.

Design context

Only the preamble is shared; the two methods' bodies are intentionally left as they are.
The base class documents "Prefer overriding simulate_expectation_values_sweep_iter", and
the sparse Simulator deliberately stays a lazy generator that streams over
simulate_sweep_iter and uses expectation_from_state_vector, while
DensityMatrixSimulator is eager and uses expectation_from_density_matrix. Collapsing the
two into one method would regress the sparse path's streaming behavior, so the helper covers
only the common setup. The helper returns the already-resolved QubitOrder so the qubit map
and the downstream simulate(...) call use the same ordering — no behavior change.

No public API change (the helper is protected); no serialization changes.

Scope / non-goals

• Does not change the eager-vs-lazy structure of either method.
• Does not touch the other Simulates* interfaces or the Clifford/MPS simulators (they
  don't override these methods).
• No new public symbols.

Test

No new behavior is introduced, so the existing expectation-value tests act as the regression
net — density_matrix_simulator_test.py, sparse_simulator_test.py, and simulator_test.py
all pass (473 tests), and they already cover the numeric expectation values, the
terminal-measurement ValueError, and the permit_terminal_measurements flag, which
exercise every line of the new helper.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 99.60%. Comparing base (b29b6d4) to head (a234874).
⚠️ Report is 19 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main    #8123      +/-   ##
==========================================
- Coverage   99.60%   99.60%   -0.01%     
==========================================
  Files        1118     1118              
  Lines      100957   100954       -3     
==========================================
- Hits       100557   100553       -4     
- Misses        400      401       +1     

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[pavoljuhas]

Thank you for the PR. The primary goal of refactoring tasks like #4209 is to make the code more readable and maintainable. I feel it does not happen for the change here. As a rough approximation the total LOC is essentially unchanged when excluding comments and docstrings and notably increased in terms of raw counts. There is an extra new indirection to follow in the simulate_expectation_values_{iter,sweep} code.

Note these methods cannot be easily replaced with a common base-class method, because one needs to call expectation_from_state_vector whereas the other expectation_from_density_matrix and they also do a different thing to get the sequence of simulation results.

This is still useful to show that #4209 should be probably closed as wontfix; it was open long enough without much complaints.