perf: optimize buffer handling and fragment gecoding

[ppmpreetham]

Improves memory efficiency and reduces CPU overhead in the WebSocket writing and fragment processing logic. Shifted to ownership transfers and in-place mutations and reduced the number of heap allocations and eliminate unnecessary buffer zero-initializations.

Operation	Before this PR	After this PR
Write Header	Allocation + Zero-init + Copy	Allocation + Copy
Decode Text	$O(n)$ Copy to new Vec	$O(1)$ Ownership transfer / Truncate

[ppmpreetham]

@littledivy , could you please check this?

[divybot]

Maintenance scan note: since this is a performance PR, it needs reproducible benchmark evidence before it is reviewable. Please add the exact commands, environment, and raw before/after numbers for the src/frame.rs write path and fragmented text path.

A small regression/perf test around fragmented text frames would also help pin that the ownership-transfer change preserves the partial UTF-8 behavior from #123. The code diff is small, but without raw numbers it is hard to tell whether the extra temporary header buffer in Frame::write is a net win across payload sizes.