ci(ric): Cut the number of pulls from AWS public ECR

[darklight3it]

Issue #, if available:

Description of changes:

The RIC PR checks pull all their images anonymously from public.ecr.aws, and they keep hitting its 429 rate limit. This PR reduces the number of pulls and how many jobs pull at the same time, without reducing what is tested.

Changes

• integration-test-matrix: max-parallel: 4 replaces the random 0–60 s jitter step. All 11 shards used to start within seconds of each other; now at most 4 run at once.
• buildspec.os.alpine.yml: 3.15 only. 3.13 and 3.14 are end-of-life and use the same musl ABI.
• buildspec.os.amazoncorretto.yml: Java 8 only. Corretto 11 is already covered by the smoke tests (test-smoke-<arch> runs amazoncorretto 11 on both archs).
• Makefile: pr-<arch> no longer runs test. The unit tests already run in the required build check (make build-<arch>). The local make pr target is unchanged.

Effect (anonymous pulls per PR, one per docker build or docker pull)

Job	Before	After	Removed
integration (11 shards)	147	111	36: alpine 3.13/3.14 and Corretto 11, 6 combos × 6 pulls
smoke (2 archs)	32	28	4: no mvn test, so no test-time JNI build
build (2 archs)	16	16	0
Total	195	155	40 (about 21%)

At most 8 jobs now pull at the same time when a PR opens, down from 15 (11 integration shards + 2 smoke + 2 build).

Each integration combo pulls binfmt, the two JNI base images (4 builds: test-time and packaging, glibc and musl) and its function image; Amazon Linux 1 skips binfmt. Many of the remaining pulls are cached manifest lookups, but the removed combos' function images (alpine 3.13/3.14, Corretto 11) were full downloads.

Coverage

Integration combos go from 21 to 15. Together with the 4 smoke combos, CI still runs every libc × arch × Java combination it ran before, apart from the two EOL Alpine versions. Each combination now runs once, except alpine 3.15/corretto11, which runs in both smoke and the integration matrix.

Cost

The integration check takes longer, because 11 shards now run in about three waves of 4. max-parallel can be raised or removed once these images come from the ECR pull-through cache (separate PR).

Followup

A followup PR will enable pull-through-cache.

Target (OCI, Managed Runtime, both): CI only, no library changes.

By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 66.66%. Comparing base (fd64949) to head (33b9bcc).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##               main     #663   +/-   ##
=========================================
  Coverage     66.66%   66.66%           
  Complexity      223      223           
=========================================
  Files            35       35           
  Lines          1020     1020           
  Branches        146      146           
=========================================
  Hits            680      680           
  Misses          288      288           
  Partials         52       52           

Flag	Coverage Δ
aarch64	66.66% <ø> (ø)
x86_64	66.27% <ø> (ø)

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Harness.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.