Add JS bindings for full libktx

[aqnuep]

This PR has been repurposed from adding the Phasmatic bindings for partial write functionality into adding bindings for nearly the full libktx functionality and a test for all the new functionality.

The PR makes the following minor API changes to libktx:

• Adds ktxTexture_GetLevelSize function.
• Makes the parameters to ktxTexture[12]_Create const.
• Adds ktxTexture2_{GetPrimaries_e, SetOETF and SetPrimaries} functions.

It adds the following functions to the JS binding while modernizing and greatly improving the binding code:

• A constructor equivalent to ktxTexture_Create.
• createCopy
• findKeyValue
• getDataSize
• getOETF
• getPrimaries
• getImage for retrieving an image from the texture. Fixes ktxTexture should expose its data via JS bindings #288.
• setImageFromMemory
• compressAstc
• compressBasis
• deflateZstd
• deflateZLIB
• addKVPairString and addKVPairByte
• deleteKVPair
• setOETF
• setPrimaries
• writeToMemory

It makes available two bindings. First is a read-only binding, libktx_read.js which has the previously available functions plus findKeyValue, getDataSize, getOETF, getPrimaries and getImage. The second, using the original name libktx.js, has functions for creating new KTX texture objects and writing them to files.

The JS ktxTexture class has been renamed texture as typical usage is with an Emscripten module instance name of ktx and new ktx.ktxTexture() looks weird. Newly added classes do not have ktx prefixes.

Compared to the Phasmatic bindings it:

• Removes compressBasisU which was bundling calls to ktxTexture2_CompressBasisEx, ktxTexture2_DeflateZstd, ktxTexture2_WriteToMemory and ktxHashList_addKVPair. Instead each function now has a binding and can be called by the JS client.
• Removes createFromBuffer which was using parameters passed from Javascript to determine the VKFormat and size of the texture then creating it with ktxTexture2_Create and finally loading an image from a passed ArrayBufferView into the texture via ktxTexture_SetImageFromMemory. Instead a ktxCreateInfo can be initialized in the JS code which can call new ktxTexture to create the texture then call setImageFromMemory to put the image into the texture object.

[MarkCallow]

I think the changes in these comments are all that are needed so ktx_js is only built when the corresponding feature is on.

[aqnuep]

Will it help to bracket these includes and any other write-related code in ktx_wrapper with#if KTX_FEATURE_WRITE?

For some reason I cannot reply on this comment, so replying here: what do you mean by that?

[MarkCallow]

@abasilak, @ViNeek please review this PR. Please pay particular attention to the following:

1. Will there be any difficulties using these bindings for the glTF-Compressor?
2. I have changed the name of the factory function for creating the ktx module from LIBKTX to createKtxModule (and createKtxReadModule). I am not planning to provide the old name as an alias because it is a simple change for users to make. I have provided an alias. What is your opinion?
3. In the test sample I am now setting window.ktx = <module returned from createKtxModule> so now I have code like var texture = new ktx.ktxTexture which is feels daft. I'd like to remove the ktx prefix from all the times that currently have it. What do you think? Note that the all the c++ code for the binding is in a ktx namespace. The extra prefixes really are superfluous in my opinion. I would, if I can figure out how, provide the old names as aliases. I have removed the ktx prefixes from the only existing class that had it, ktxTexture, and from the new classes and have provided an alias from ktxTexture to texture.
4. Point out any errors or places for improvement in the code.
5. Comment on the naming style of the classes, enums and values in the JS API.

I would like the glTF-Compressor to be updated to use this binding. Is there any room in your existing Khronos contract for that work? Better reply to me privately on this question.

[MarkCallow]

Here is a screenshot of the test for the new features.

[ViNeek]

Hi @MarkCallow thank you for the great work.

Regarding you comments:

1. I can' t see any difficulties in integrating these changes to the gltf Compressor.
2. This looks nice.
3. Removing the prefix seems like a good, clean choice.
4. The code seems very well structured and documented.
5. I really like the ktx.SYMBOL style.

Test page also looks very good.

[MarkCallow]

@ViNeek thank you for your review and comments. I have made one final (fingers crossed) change. I have changed the enumerator names to be the same as the libktx names minus the ktx{,_} prefixes and _e suffixes for consistency and to make it easier for people to recall what the JS name will be. Please take a look at 6966f6c.