Pyembroidery

Defective DST headers not in ascii would cause a critical error preventing the body from being read. Properly this should merely prevent the header from being read.

There are three strict color only file formats in `COL`, `INF`, `EDR` which provide no stitch data but do provide color information. These now have writers and readers and are processed as type `color`. This is useful if loading a DST with an EDR file. In addition, I added JSON objects. These allow the writing and reading of the patterns in a mostly lossless JSON file (some bytearray info won't serialize). This should also be useful for a variety of purposes where data interchange matters.

* This included a small tweak to allow writing in text mode (json required it).
* A bug fix to EmbCompress for 3.6 division.
* A few tweaks based on machine tests of VP3 file format.

Adds a reader for HUS format along with the required decompression format written from scratch in pure python.

This includes a fix for issue 77.

Some corrections to VP3 correction.
* Add an END command at the pattern end.
* Avoid double trims for explicit and implicit trims.
* Do not treating an END command as a TRIM.

---

This also includes several minor corrections of `is` vs `==` and modifies the EmbEncoder to pass along some higher level bits to the writers. This should generally have no effect, but, could be of some utility in writers where custom higher level bit flags served some purpose. Currently VP3 will deal with a ALTERNATIVE flag on STITCH to write a long form stitch, for testing purposes.

VP3 trims are signalled as `0x80 0x03`. I previously viewed this as an END. This resulted in two effects. When trims were loaded it would force end a block read which would also skip appending the color change which caused it to garble. And I didn't read write trims as anything.

This adds a few more read interpolations for PES and PEC that turn duplicate color color_changes into STOP commands, and vice versa. These are done natively for the PES and PEC formats. This makes STOP a fully qualified command in the format.

Changes were made to how options are listed in the supported formats. This could be breaking in a rare edge case.

The major changes to the scheme is that `PyEmbroidery. supported_formats()` now includes `versions` which can be requested through settings `{ "version": <request> }`, this is a fundamentally more consistent methodology especially for some writers that may require a lot of different versioning information. Many of the options were cryptic and would need to be configured in a strange ways. With versions you can simply list the versions and write them out. There is now an `extensions` key which provides a tuple of different extensions that are common for the reader to load. This differs from `extension` which provided a single typical export type.

There are interpolation corrections for jef and dst with regard to trims, after the files are loaded. This should largely remove the need to call `get_pattern_interpolate_trim()` as a routine operation.