Reword "computationally heavy" language in docs

[sffc]

Our docs have verbiage such as:

For that reason, one should think of the process of formatting a date in two steps - first, a computational heavy construction of DateTimeFormatter, and then fast formatting of DateTime data using the instance."

This is somewhat misleading. We should come up with better verbiage.

[zbraniecki]

Do we have any measurements of the actual cost of DTF construction?

[sffc]

Let's block this on #3865 and write the verbiage based on measurements after those changes land.

[Manishearth]

Assigning to @sffc since it's a part of the whole set of datetime changes

[sffc]

This language no longer exists in the current docs, and my new docs discuss various tradeoffs, although not specifically this one. The principle of "expensive constructor and fast format function" is one that would apply horizontally across ICU4X. So I will close this issue.