docs: use SciMLLogging.None() instead of verbose=false (OrdinaryDiffEq v7)

[ChrisRackauckas-Claude]

Problem

The Documentation build on main is red. The @example blocks in docs/src/nnblock.md and docs/src/symbolic_ude_tutorial.md pass verbose = false (a Bool) to solve. OrdinaryDiffEq v7 no longer accepts a Bool for verbose:

ArgumentError: Passing a `Bool` for `verbose` is no longer supported in OrdinaryDiffEq v7.
Use `DEVerbosity()` or a preset like `Standard()`, `None()`, etc. from SciMLLogging.

This caused failed to run @example block errors at nnblock.md:208-292 and symbolic_ude_tutorial.md:74-83, which then cascaded into UndefVarErrors (new_sol1, opt_sol, oprob_fitted) in downstream blocks and terminated the build ([:example_block]).

Fix

verbose = false means "suppress all solver output", which maps to the SciMLLogging.None() preset (per DiffEqBase, _process_verbose_param(::Val{false}) resolves to DEVerbosity(SciMLLogging.None())). This PR:

• Adds SciMLLogging (compat "2") to docs/Project.toml (resolves to the same v2.0.0 already in the docs Manifest, no other resolution changes).
• Replaces verbose = false with verbose = SciMLLogging.None() in both tutorials, and adds import SciMLLogging to the relevant example modules.

Verification (local, Julia 1.10 / docs env)

• Confirmed solve(prob, Tsit5(); verbose = SciMLLogging.None()) returns retcode = Success, while verbose = false still throws the ArgumentError.
• Ran julia --project=docs docs/make.jl: the build now evaluates the nnblock.md potplate block (previously the first failure) with zero ArgumentError/failed to run/UndefVarError markers, and proceeds into the subsequent new_sol1 and equation_search blocks. (The local run was OOM-killed at the memory-heavy SymbolicRegression block on a heavily-loaded shared machine; this is unrelated to the docs error and will not occur on CI runners.)
• Independently reproduced the symbolic_ude_tutorial.md loss-block solve call with verbose = SciMLLogging.None(): retcode = Success.

Note

There is a separate, pre-existing, platform-specific failure on the tests / All (julia lts, macos-latest) job only: @test res2.objective < 1.5e-4 evaluates to 0.000270... < 0.00015 on macOS aarch64 LTS. It passes on ubuntu/windows for both lts and 1. This is optimizer-convergence noise (LBFGS/Adam landing in a slightly worse local minimum due to aarch64 BLAS/LAPACK differences) and is intentionally not touched here — loosening that tolerance would risk masking a real regression on other platforms. Out of scope for this PR.

Please ignore until reviewed by @ChrisRackauckas