Modernize Julia CI and compat floor

[bernalde]

Summary

• Raise the advertised Julia compatibility floor from 1.9 to 1.10.
• Update compat to the modernized stack: QUBODrivers 0.4 and QUBOTools 0.12.
• Add a Julia 1.10 CI lane while keeping latest stable Julia via version: '1'.
• Add a metadata regression test for the compat and CI lanes.
• Add changelog notes for the Julia floor and dependency support changes.

Tests run

• julia --startup-file=no -e "import Test, TOML; root = pwd(); project = TOML.parsefile(joinpath(root, \"Project.toml\")); compat = project[\"compat\"]; Test.@testset \"Compatibility metadata\" begin; Test.@test compat[\"julia\"] == \"1.10\"; Test.@test compat[\"QUBODrivers\"] == \"0.4\"; Test.@test compat[\"QUBOTools\"] == \"0.12\"; ci = read(joinpath(root, \".github\", \"workflows\", \"ci.yml\"), String); Test.@test occursin(r\"version:\\s*'1\\.10'\", ci); Test.@test occursin(r\"version:\\s*'1'\", ci); end": passed, 5/5.
• julia --startup-file=no --project=/tmp/mqlib-issue5-test.MadlRy -e 'using Pkg; Pkg.test()': passed on Julia 1.10.11 from a clean temporary copy without the ignored local Manifest.toml; resolved QUBODrivers 0.4.0 and QUBOTools 0.12.0.
• bash test/native_c_api_smoke.sh: passed; both C ABI smoke binaries returned objective 6.
• git diff --check: passed.

Notes

• No docs project/workflow, lint command, type-check command, or formatter command is documented in this repository.
• The package test was run from a temporary copy so the ignored local Manifest.toml with development paths would not affect dependency resolution.

Closes #5

[bernalde]

No blocking issues found. I reviewed the PR diff, surrounding compat/test/CI conventions, and documentation surface. The compat floor, dependency bounds, CI matrix, changelog note, and metadata tests are consistent with issue #5, and local plus CI validation are passing.\n\nI attempted to submit this as APPROVE, but GitHub rejected it because the authenticated account is the PR author: Review Can not approve your own pull request. Recording the review as COMMENT for that reason.