Simplify realloc logic

[febo]

Problem

The current way that realloc tracks the original data length is unnecessarily complex.

Solution

This PR simplifies the logic on realloc to store difference between the original data length and the current length.

[rustopian]

Looks good overall, but I do wonder about a possible exploit of multiple realloc() calls given the limitations of the delta approach - see comment

[ilya-bobyr]

Using zero initialization to both avoid length assignment and make the logic simpler is really smart :)

The base branch was changed.

[joncinque]

The change looks good to me!

There is an edge case if someone calls close() and then resize() within the same instruction.

With the current code, resize() would incorrectly return an error for going over the limit, even if the account previously had some space.

How about setting the difference during close()? close_unchecked() can stay as it is, but it would be good to add a comment saying to avoid resizing afterwards.

[febo]

The change looks good to me!

There is an edge case if someone calls close() and then resize() within the same instruction.

With the current code, resize() would incorrectly return an error for going over the limit, even if the account previously had some space.

How about setting the difference during close()? close_unchecked() can stay as it is, but it would be good to add a comment saying to avoid resizing afterwards.

Good point! Fixed close() as suggested.

[joncinque]

Looks great to me!