[cgroup] fix uint32 overflow in CPU quota computation

[ManoharPaturi]

Problem

cgroup_cpu_ms_per_sec is a uint32 config field, and both quota writers compute the microsecond value with * 1000U, which evaluates in 32 bits. Any configured value ≥ 4294968 wraps around and writes a completely unrelated limit to the kernel:

configured ms/s	intended µs	written µs (32-bit wrap)
4,294,968	4,294,968,000	704
5,000,000	5,000,000,000	705,032,704

So an operator requesting ~4300 CPU-seconds per second silently gets a 704 µs quota in cgroup v1, and cgroup v2 gets an arbitrary wrapped cpu.max value — the enforced limit has no relation to the configuration.

Change

Compute the quota in uint64_t in both writers (cgroup.cc for v1, cgroup2.cc for v2). No behavior change for values that don't overflow.

Testing

Built and ran the full test suite (unit + cmdline) plus an ASan/UBSan build on Ubuntu 24.04 — all green, no new findings. The overflow itself is demonstrated by the table above (plain uint32 arithmetic).

[ManoharPaturi]

Small branch hygiene fix: my local CI workflow file accidentally ended up in this branch — force-pushed the branch rebuilt from current master with only the intended change. The diff is now just the patch itself; the actions scan failure from earlier is gone.

[ManoharPaturi]

@robertswiecki Could you take a look when you have a moment? This fixes a uint32 overflow in the CPU quota computation: cgroup_cpu_ms_per_sec * 1000U evaluates in 32 bits, so values above 4,294,967 wrap (e.g. 4294968 ms/s silently becomes a 704 µs quota). Both the v1 and v2 writers are affected; the fix computes in uint64_t. Verified with a full build, the test suite, and an ASan/UBSan run.