SG-4024: Add MAX_RUNNERS constraint.

[hllvc]

Adds a MAX_RUNNERS constraint to the sg-runner-autoscaler to cap concurrent runner provisioning. This aligns with StackGuardian’s ongoing capacity and cost controls for CI runner usage under SG-4024.

Changes

• MAX_RUNNERS config: Introduces a new constraint to limit the total number of concurrently provisioned runners; acts as a hard ceiling in scaling decisions.
• Scaling logic enforcement: Validates desired scale against MAX_RUNNERS before provisioning or queuing new runners, preventing burst-induced over-allocation.
• Behavioral defaults: When MAX_RUNNERS is unset, autoscaler behaves as before (no cap); when set to an integer, the ceiling is enforced consistently.
• Edge-case handling: Ensures correct behavior under job bursts, cooldowns, and queue backpressure; avoids thrashing at the boundary.
• Configuration surface: Adds parsing/support for configuration (env/params) so ops can tune caps per environment/org.
• Refactor/cleanup: Minor adjustments around the scaling path to centralize cap checks and reduce duplication.
• Scope of diff: +43 / −16 in one file; focused change set to introduce and enforce the constraint.

Testing

No tests or manual verification are included in the PR as visible here.

[notion-workspace]

Update Autoscaler to support MAX_RUNNERS

[taherkk]

if this is the case you should exit immediately. This invalid configuration

[taherkk]

exit instead of warning