Reuse buffer to avoid memory spikes

[akirayamamoto]

There is a breaking change in this PR. The behavior has not changed but one method needs more parameters. Data generated by the previous version is compatible with this version.

We improved the use of memory to avoid memory spikes and pod restarts in Octopus server.

For each chunk of data that we need to copy, instead of allocating a new array of bytes we will re-use the existing buffer.

Memory allocated before 102 GB. After 8 MB. See details below.

[sc-25897]

Before

Method	Mean	Error	StdDev	Gen0	Gen1	Gen2	Allocated
ApplyBigDelta_Different	16.19 ms	0.311 ms	0.576 ms	375.0000	375.0000	375.0000	50 MB
ApplyBigDelta_Identical	2,066.69 ms	41.238 ms	98.802 ms	363000.0000	363000.0000	363000.0000	102400.77 MB

Method	Mean	Error	StdDev	Gen0	Gen1	Gen2	Allocated
ApplyBigDelta_Different	16.36 ms	0.279 ms	0.261 ms	375.0000	375.0000	375.0000	50 MB
ApplyBigDelta_Identical	1,983.46 ms	37.114 ms	93.113 ms	364000.0000	364000.0000	364000.0000	102400.77 MB

After

Method	Mean	Error	StdDev	Gen0	Gen1	Gen2	Allocated
ApplyBigDelta_Different	4.577 ms	0.0874 ms	0.0935 ms	78.1250	78.1250	78.1250	8 MB
ApplyBigDelta_Identical	4.055 ms	0.0388 ms	0.0344 ms	62.5000	62.5000	62.5000	8 MB

Method	Mean	Error	StdDev	Gen0	Gen1	Gen2	Allocated
ApplyBigDelta_Different	4.319 ms	0.0842 ms	0.1475 ms	78.1250	78.1250	78.1250	8 MB
ApplyBigDelta_Identical	3.978 ms	0.0265 ms	0.0235 ms	62.5000	62.5000	62.5000	8 MB

[shortcut-integration[bot]]

This pull request has been linked to Shortcut Story #25897: Delta package upload could be causing memory spikes which might cause the pod to restart.