`AWS_MAX_ATTEMPTS` has no effect with v2

Confirm by changing [ ] to [x] below to ensure that it's a bug:
- [x] I've gone though the [User Guide](https://docs.aws.amazon.com/cli/latest/userguide/cli-chap-welcome.html) and the [API reference](https://docs.aws.amazon.com/cli/latest/reference/)
- [x] I've searched for [previous similar issues](https://github.com/aws/aws-cli/issues) and didn't find any solution

**Describe the bug**

With the v2 CLI, the `AWS_MAX_ATTEMPTS` environment variable does not change the number of max attempts. The variable is documented to work: . It does work with the v1 CLI.

**SDK version number**

aws-cli/2.0.46 Python/3.8.5 Linux/5.4.69 source/x86_64.nixos.20

**Platform/OS/Hardware/Device**

NixOS 20.09

**To Reproduce (observed behavior)**

1. Start an S3 upload like `AWS_MAX_ATTEMPTS=10000 aws s3 cp --debug /tmp/randfile s3://my-bucket/test`
2. Disable networking
3. After about 5 minutes, observe "botocore.retries.standard - DEBUG - Retry needed but retry quota reached, not retrying request."

**Expected behavior**

Upload should continue to retry until `AWS_MAX_ATTEMPTS` attempts are made.

**Logs/output**

[aws-s3-cp.log](https://github.com/aws/aws-cli/files/5405259/aws-s3-cp.log) for `AWS_MAX_ATTEMPTS=10000 aws s3 cp --debug /tmp/randfile s3://my-bucket/test > aws-s3-cp.log 2>&1`.
See "To Reproduce" for the relevant entry.

Start by reproducing the documented AWS_MAX_ATTEMPTS=10000 command and reviewing the attached aws-s3-cp.log, especially the retry-quota message. Compare the v2 behavior with the stated v1 behavior and the linked configuration documentation. Done means the v2 S3 upload honors the configured maximum number of attempts.

Written by the indexing model from the issue text.