[#2783] fix(spark): correct compression ratio and failure counters in Spark UI

[wForget]

What changes were proposed in this pull request?

• Calculate the Spark UI compression ratio using floating-point division.
  • Return 0.0 when the compressed shuffle size is zero.
  • Convert uncompressedShuffleBytes to Double before division.
• Initialize failedTaskNumber and failedTaskMaxAttemptNumber to 0 instead of -1.

Why are the changes needed?

The compression ratio was calculated by dividing two Long values, so Scala performed integer division and discarded the fractional part.

For example, 150 / 100 was calculated as 1 and displayed as 1.0 instead of the expected 1.5.

Additionally, initializing the failure count to -1 caused the first failure to increment the count to 0, resulting in an incorrect failure summary. Starting the failure-related fields at 0 correctly represents the initial state and counts the first failure.

Fix: #2783

Does this PR introduce any user-facing change?

Yes.

• Spark UI now displays fractional compression ratios correctly.
• Shuffle failure counters start at zero and include the first failure.

How was this patch tested?

[github-actions]

Test Results

 3 829 files   - 4   3 829 suites   - 4   8h 1m 5s ⏱️ - 10m 10s
 1 266 tests ±0   1 255 ✅ ±0  11 💤 ±0  0 ❌ ±0 
19 149 runs   - 7  19 101 ✅  - 7  48 💤 ±0  0 ❌ ±0 

Results for commit 10b0f10. ± Comparison against base commit b1ac32f.

♻️ This comment has been updated with latest results.