fix: read a large range without error and add test

[Ranxy]

I hereby agree to the terms of the CLA available at: https://databend.rs/dev/policies/cla/

Summary

Summary about this PR
Close #943 fix error if block range read with a large range and add test for it.

[Xuanwo]

Thanks for the contribution! Mostly LGTM, but I prefer to make a new test case so that we can address them more easier.

[Xuanwo]

[2022-12-12T09:54:23Z DEBUG opendal::services] service=s3 operation=read path=ffbd8eab-4916-4b12-b302-a4a4387b7ada.gz range=0- -> got reader
[2022-12-12T09:54:23Z DEBUG opendal::services] service=s3 operation=read path=d2fba631-9198-49c6-b55f-f64bcfd468c8 range=1139367-18446744073709551614 -> got reader
[2022-12-12T09:54:23Z DEBUG opendal::services] service=s3 operation=read path=d2fba631-9198-49c6-b55f-f64bcfd468c8 has_read=2034952 -> dropped reader
thread 'services_s3_write::test_read_large_range' panicked at 'assertion failed: `(left == right)`
  left: `2034952`,
 right: `895585`: read size with large range', tests/behavior/write.rs:425:5

Interestingly, minio failed this test case. But minio should support range requests. We need more investigating.

Can you try use smaller size like u32::MAX?

[Xuanwo]

minio will parse range into int64:

https://github.com/minio/minio/blob/444ff20bc5268d61b329c1d1bc7eb22aeb5f96de/cmd/httprange.go#L106-L164

[Ranxy]

Okay, let me try.

[Ranxy]

It seems that tests on hdfs are not stable, fortunately, the minio problem is solved.
Can we restart this test separately, Or just push an empty commit to restart all?

[Xuanwo]

It seems that tests on hdfs are not stable

It's a known issue about hdfs's test: #465

And I have no idea on this, let's just retry it.

[Xuanwo]

Thanks for contribution!