[jbonofre]

What's Changed

Follow-up to #1231. ClobConsumer reads a CLOB in chunks of 256 characters and encodes each chunk to UTF-8 on its own, so a character outside the BMP split between two chunks was encoded as ??.

• A chunk that ends with a high surrogate, while characters remain, now leaves it for the next chunk, so the pair is encoded together.
• The read position advances by the length of the chunk instead of the requested size. The shorter chunk needs this, and it also stops skipping characters when a driver returns fewer than requested.
• An empty chunk before the end of the CLOB throws an SQLException instead of looping forever.

ClobConsumerTest covers emojis across chunk boundaries for both consumers. The test fails without the fix: 255 a followed by 😀 read back as 257 bytes instead of 259.

This changes other lines of ClobConsumer than #1479, and the two branches merge cleanly.

Closes #1478.