PARQUET-866: API fixes for ARROW-33 patch

[wesm]

See ARROW-33 patch apache/arrow#322

@xhochy this fails on Int96 timestamps. I'm not sure why yet

[xhochy]

+1, LGTM.

One thing that is different between ColumnReader::Impl::ReadNonNullableBatch<INT96> and the general case is that in the general implementation we use out_ptr + valid_bits_idx_ and in ReadNonNullableBatch<INT96> only out_ptr_

[wesm]

per discussion in apache/arrow#322, I think the Int96 thing is a bug that was exposed by the refactoring. if you could take a look and let me know if you agree

[wesm]

here's the view from inside gdb (at arrow/compare.cc:348)

(gdb) p left_data
$1 = (const uint8_t *) 0xcd39c0 "/\263\256\243\201t"
(gdb) p right_data
$2 = (const uint8_t *) 0xcd2840 "/\263\375\064", <incomplete sequence \303>

The first byte is equal, and because of the bug that was the only byte being compared amazingly.

[wesm]

OK, fixed the off-by-one error, the build should pass now

[wesm]

build passed on my fork: https://travis-ci.org/wesm/parquet-cpp/builds/198973522. Merging