Revert "Revert "[Client] chunked get requests (#22455)""

[ckw017]

Why are these changes needed?

Reverts #22455

PR was originally reverted because it coincided with timeouts in test_remote_package_uri. I suspect the originally PR wasn't the cause of flakiness, since the only parts that this PR could affect in that test is in the codepath for synchronous gets, and it seems unlikely that breaking synchronous gets would only break a single test.

Tested this revert multiple times against CI in #22713 and was unable to reproduce flakiness.

To be extra careful, added two changes (diff can be seen here):

• Tighten the error handling in the proxier to cover case where an exception occurs while yielding from the raylet server stub
• Exit early in _get_object_iterator once all chunks are yielded

Related issue number

Checks

• I've run scripts/format.sh to lint the changes in this PR.
• I've included any doc changes needed for https://docs.ray.io/en/master/.
• I've made sure the tests are passing. Note that there might be a few flaky tests, see the recent failures at https://flakey-tests.ray.io/
• Testing Strategy
  • Unit tests
  • Release tests
  • This PR is not tested :(

[jjyao]

Just curious, if this PR was not the root cause, do we know the actual cause of the flakiness?

[ckw017]

We'll find out if this PR really was the cause flakiness again soon, but if it turns out it wasn't this PR, then I suspect the cause would have been some kind of temporary degradation/outage with github, since the params for the timed out test had "https://github.com/shrekris-anyscale/test_module/archive/HEAD.zip" for the remote uri

[mwtian]

test_runtime_env_working_dir_remote_uri looks to have the same flakiness before and after the PR.