feat: Add support for {DataFrame,LazyFrame,Series}.clear method

[FBruzzesi]

What type of PR is this? (check all applicable)

• 💾 Refactor
• ✨ Feature
• 🐛 Bug Fix
• 🔧 Optimization
• 📝 Documentation
• ✅ Test
• 🐳 Other

Related issues

• Closes enh: Add support for {DataFrame,LazyFrame}.clear #2890
• Closes [Enh]: Add narwhals.empty_like or another way to construct an empty Narwals frame with known schema and implementation #3296
• feat(python): allow for simple creation of n-row empty frame/series via clear pola-rs/polars#7095

Checklist

• Code follows style guide (ruff)
• Tests added
• Documented the changes

If you have comments or can explain your changes, please do so below

Opening as draft since we need to decide what to do for the cases in which the original dtype is non nullable for pandas (see #2890 (comment))

[dangotbanned]

I hope its okay I linked the polars PR that added the null behavior.

I was looking to see it came from a feature request or had an origin in pandas for (#2890 (comment)) - but the search continues! 😅

[FBruzzesi]

As pointed in the commit message "keep it simple for now", for now I am failing the case of n>0 for both pandas-like and dask. All other backend should be ok to review for now .

Update: 3006fec is an attempt for pandas-like backends that allows to clear(n) only if it does not change the underlying dtype.