Slotted array faster defragmentation

[Scooletz]

This PR changes the SlottedArray.Defragment method that is clearly visible in traces and previously was copying data twice, once from the array to a buffer, then buffer to an array. This PR removes the need of the auxiliary buffer which required renting & releasing but also introduced one more copy.

The benchmarks below does not show the gains that are observed. P95 of the block application to the database dropped by ~20% which is a lot.

Benchmark

Before

Method	Mean	Error	StdDev	Code Size
Defragment_first_key_deleted	372.76 ns	3.585 ns	3.178 ns	7,438 B
Defragment_last_key_deleted	30.19 ns	0.473 ns	0.443 ns	5,609 B

After

Method	Mean	Error	StdDev	Code Size
Defragment_first_key_deleted	258.77 ns	2.896 ns	2.567 ns	5,553 B
Defragment_last_key_deleted	29.75 ns	0.274 ns	0.256 ns	5,607 B

[github-actions]

Package	Line Rate	Branch Rate	Health
Paprika	85%	80%	✔
Summary	85% (4194 / 4925)	80% (1347 / 1682)	✔

Minimum allowed line rate is 75%