[pull] master from mozilla:master - #509

…etBytes`

The `getBytes` methods will *always* return a Uint8Array[1], and remember that even an empty TypedArray is truthy (i.e. `!!new Uint8Array(0) === true`).
Hence we can remove a few pointless checks from the `DecryptStream` and `RunLengthStream` classes, which should improve code readability a tiny bit.

---

[1] See [stream.js](https://github.com/mozilla/pdf.js/blob/e051279385c5cf141935037b340e757081052990/src/core/stream.js#L45-L51), [chunked_stream.js](https://github.com/mozilla/pdf.js/blob/e051279385c5cf141935037b340e757081052990/src/core/chunked_stream.js#L184-L193), and [decode_stream.js](https://github.com/mozilla/pdf.js/blob/e051279385c5cf141935037b340e757081052990/src/core/decode_stream.js#L86-L110).

…ecompiling font-data

This avoids creating "duplicate" `TextDecoder`/`TextEncoder` instances.
For example, when rendering all 1310 pages of the `pdf.pdf` document (from the test-suite) this patch reduces the number of `TextDecoder`/`TextEncoder` instances from `73` *each* to just a single one each.

*Note:* Compared to PR 21801 the reduction is obviously much smaller, however it still seems worthwhile to avoid unnecessary object creation.

Remove a few unnecessary checks after calling `BaseStream.prototype.getBytes`

…TextDecoder-TextEncoder

Cache the `TextDecoder`/`TextEncoder` instances used when compiling/decompiling font-data

Vertically center the text in single-line and comb text fields (bug 2055455)