perf: add missing std::moves to save refcounting

[eric-wieser]

I haven't benchmarked this.

[leanprover-community-bot]

Mathlib CI status (docs):

• ❗ Batteries/Mathlib CI will not be attempted unless your PR branches off the nightly-with-mathlib branch. Try git rebase 4077bf2c05b74a2eaaf1db9be957236f29614f63 --onto 4ce7ad19cea348506876685a19a7e0cca442c6d0. You can force Mathlib CI using the force-mathlib-ci label. (2025-10-17 12:51:37)

[leanprover-bot]

Reference manual CI status:

• ❗ Reference manual CI will not be attempted unless your PR branches off the nightly-with-manual branch. Try git rebase 4077bf2c05b74a2eaaf1db9be957236f29614f63 --onto d3dda9f6d4428a906c096067ecb75e432afc4615. You can force reference manual CI using the force-manual-ci label. (2025-10-17 12:51:38)

[Kha]

!bench

[leanprover-radar]

Benchmark results for 09b2bfc against 4077bf2 are in! @Kha

[leanprover-bot]

Here are the benchmark results for commit 09b2bfc.
There were significant changes against commit 4077bf2:

  Benchmark                  Metric                    Change
  =====================================================================
+ bv_decide_large_aig.lean   branch-misses              -4.3% (-47.6 σ)
- stdlib                     instantiate metavars        1.1% (148.3 σ)
- stdlib                     process pre-definitions     1.4%  (63.6 σ)

[Kha]

@eric-wieser I'm a bit more hesitant here because because incorrect use of std::move can easily lead to bugs without great static analysis around it. It still shouldn't be UB, unless of course there are other bugs in e.g. move constructors. But the benchmarks don't seem to justify this extra needed care.