Check that all options bags are taken by value

[sffc]

Follow-up to discussion in #1077 to verify that we take arguments by value where appropriate and by reference where appropriate, with a focus on options bags.

[sffc]

I scope-creeped this to look at all functions that take by value or reference.

There are a number of places where we take Copy types by reference, especially by self reference:

• try_add and try_sub and number on Copy types such as IsoHour, NanoSecond, etc: should it take self or &self? Currently it takes &self
• AnyCalendarKind::as_bcp47_*: it takes a &self but should it take self since it is Copy?
• CodePointInversionListBuilder::*_range_*: it takes &impl RangeBounds but should it take impl RangeBounds?
• Also icu::collections::codepointinvlist::deconstruct_range
• The icu_locid Copy subtags have as_str, strict_cmp, and normalizing_eq; should they take self?
• icu::properties::maps::CodePointMapDataBorrowed and similar classes are Copy; should their methods take self?
• GmtOffset functions
• DataKey, DataKeyPath, DataKeyHash, and DataKeyMetadata functions

Thoughts on the above? @Manishearth

[Manishearth]

I would default all of those to self, except the ones below

CodePointInversionListBuilder::*_range_*: it takes &impl RangeBounds but should it take impl RangeBounds?

nope, RangeBounds doesn't require Copy and the main Range types aren't Copy

The icu_locid Copy subtags have as_str, strict_cmp, and normalizing_eq; should they take self?

No strong opinion here, really, probably prefer self as a general style rule

[sffc]

I'm making a PR for this issue and opening #2637 for utils crates.