[Bug] skipFully spins forever on a zero-byte skip at end of stream

[LuciferYang]

Bug description

HadoopSeekableInputStream.skipFully (and its copies in every HadoopCompliantFileIO wrapper: azure, cosn, gs, jindo, obs, oss, s3) loops with

while (bytes > 0) { bytes -= in.skip(bytes); }

A blocking stream at end of file returns 0 from skip, so a small forward seek — the vectored-read seek helper skips gaps of up to 1 MiB instead of seeking — past a truncated or shorter-than-expected file spins the loop forever instead of failing the read.

Expected behavior

Throw EOFException when a skip call makes no progress, like Hadoop's own IOUtils.skipFully treats premature EOF.