docs(readme): record colour handling, and the new 4:2:0 10-bit format

[StuartCameronCode]

Three gaps between the README and what 1.0 actually ships.

• Output colour format was missing 4:2:0 10-bit (added in fix: NVENC and QSV cannot encode 4:2:2 (#74) #76). The row now also says why it matters beyond playability: no NVIDIA card before the RTX 50 series can encode 4:2:2 at all, so for a 10-bit source this is the option that keeps the precision and still encodes.
• Hardware encoding now says a format the GPU can't take is converted rather than failing the job, and that the panel says which and why first.
• Colour tagging was undocumented entirely — despite being one of the larger behavioural fixes in this release. Without it every output was untagged and therefore read as BT.601 limited, shifting the colours of any BT.709 or full-range source. That, the per-pass depth handling, and the 8-bit-units convention behind every threshold in the interface now have a <details> block, covering the same ground as the new in-app explanation.

The filter table itself was already current — #75 refreshed it, and #76 added no passes.

Docs only, so ci-test.yml's paths-ignore skips the test gate.