fix(output): improve violation comprehension in CI/CD mode

[B1naryVision]

Problem

When running with the CI/CD flag (-j), a single rule breaching its
cicdmaxissues threshold caused ASIST to print findings for all rules.
With e.g. 4 rules at 10 issues each and a cap of 10, introducing 2 new
violations printed ~40 findings, making the 2 new ones very hard to spot.

Additionally, findings JSON was written to stdout while the threshold
summary went to stderr, so output ordering was unpredictable across
platforms and CI log viewers.

Changes

• CI/CD filtering — when thresholds are breached, the findings JSON now
  contains only the rules that exceeded their cicdmaxissues limit
  (Count reflects the filtered total). Exit code behavior is unchanged.
• Single stream, fixed layout — findings and the threshold summary are
  both written sequentially to stdout: issue details on top, summary at the
  very bottom.
• Explicit pass message — CI/CD runs with no breaches now end with
  No rules exceeded their cicdmaxissues threshold. instead of silence.
• Threshold logic split into countFindingsPerRule, getViolatedRuleIds
  and filterFindingsByRules helpers; CheckThresholdViolations keeps its
  existing signature and output format.

No regression without the flag

Runs without -j are byte-for-byte unchanged: all findings printed, no
threshold text, exit 0.

Testing

• New unit tests for breached-rule selection/sorting, filtering + count
  update, and the empty-rule-list edge case; full go test ./... passes.
• Verified end-to-end against integrationtests/src:
  • -j + breach → only breached rule's findings, summary at bottom,
    stderr empty, exit 1
  • -j + no breach → all findings + pass message, exit 0
  • no flag → unchanged output, exit 0