Preserve millisecond precision during serialization

[richardwang1124]

The Ruby SDK currently truncates timestamps to whole seconds. This PR preserves millisecond precision on timestamps to align with other SDKs.

---

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

1. To make sure we include your contribution in the release notes, please make sure to add description entry for your changes in the "unreleased changes" section of the CHANGELOG.md file (at corresponding gem). For the description entry, please make sure it lives in one line and starts with Feature or Issue in the correct format.

2. For generated code changes, please checkout below instructions first:
   https://github.com/aws/aws-sdk-ruby/blob/version-3/CONTRIBUTING.md

Thank you for your contribution!

[jterapin]

LGTM - I would suggest that you want for the catapult pr build to finish before merging.

[github-actions]

Detected 7 possible performance regressions:

• aws-sdk-cloudwatch.client_mem_allocated - z-score regression: 0.61 -> 0.62. Z-score: 27.53
• aws-sdk-cloudwatch.put_metric_data_small_allocated_kb - z-score regression: 47.97 -> 48.01. Z-score: Infinity
• aws-sdk-cloudwatch.put_metric_data_large_allocated_kb - z-score regression: 384.11 -> 384.54. Z-score: Infinity
• aws-sdk-cloudwatchlogs.client_mem_allocated - z-score regression: 0.62 -> 0.62. Z-score: 27.77
• aws-sdk-dynamodb.client_mem_allocated - z-score regression: 0.65 -> 0.65. Z-score: 27.77
• aws-sdk-s3.client_mem_allocated - z-score regression: 1.02 -> 1.02. Z-score: 21.36
• aws-sdk-sns.client_mem_allocated - z-score regression: 0.61 -> 0.61. Z-score: 28.38