[Forked] HADOOP-19027. S3A: S3AInputStream doesn't recover from channel exceptions

[shkhrgpt]

Original Description:

This is a fraction of apache#6425 backported to the V1 SDK, which does a lot of V2-specific translation/unwinding.

416 responses are mapped to RangeNotSatisfiableEOFException, whose retry policy is: fail.

Classic EOFException is now retried as connection failure. calls to read() and lazyseek all retry on this with full retry policy, including handling of socket errors.

Everything related to new sdk, its complex http channel stuff: omitted.
This means that the following will be handled

• EOF on socket read
• connection reset on socket read

Out of range GET is mappped to RangeNotSatisfiableEOFException so it is not retried,
this is critical. Without it ITestS3AOpenCost.testReadPastEOF() will time out as it is continually retrying.

How was this patch tested?

Cherrypick of new ITestS3AOpenCost.

For code changes:

• Does the title or this PR starts with the corresponding JIRA issue id (e.g. 'HADOOP-17799. Your PR title ...')?
• Object storage: have the integration tests been executed and the endpoint declared according to the connector-specific documentation?
• If adding new dependencies to the code, are these dependencies licensed in a way that is compatible for inclusion under ASF 2.0?
• If applicable, have you updated the LICENSE, LICENSE-binary, NOTICE-binary files?

[violetnspct]

Hello! I’m Violet Nspct, your AI-powered test review assistant.

I can help improve your pull requests by suggesting missing or additional tests. Use the command below:

• @violetnspct test → I’ll analyze your PR and recommend test cases you may want to add.

I’m here to make your code more reliable—just let me know when to start!

[shkhrgpt]

@violetnspct test