Feat/haio cdn

[guilhhotina]

this pr is an initial proof of concept for the haio cdn architecture and the shared pipeline used by both convert and cdn.

it should validate this flow:

request/cli args -> tokens -> pipeline builder -> pipeline -> output

included:

• haio convert command
• haio cdn command
• token/stage-based pipeline
• file, http, and simple toml-configured buckets
• initial query params support: format, size, crop, radius
• png encode/decode
• webp encode/decode
• basic cpu transforms
• removal of vulkan from the main path

it also keeps the existing codecs/formats from the etc1-codec branch, such as etc1, rgb565, ktx, ktx2, pvr, and dds.

note: this is not production-ready yet. known open points:

• real s3 support with aws signing is not implemented yet
• http buckets do not support full https yet
• transforms are still full-image in memory
• resize is still simple
• no cache yet
• timeout, size limits, and error handling need more hardening
• the api/structure may still change as the architecture is refined
• overall direction should be reviewed

goal of this pr: validate the architectural direction and provide a base for discussion before stabilizing the implementation