ENH: Migrate xsf header files from scipy repo to xsf repo

[steppi]

This PR migrates header files (and .clang-format) from scipy/scipy/special/xsf to xsf/include/xsf while preserving git history.

I used git-filter-repo as described here, #1 (comment), but in order to preserve history across the rename from scipy/scipy/special/special to scipy/scipy/special/xsf here https://github.com/scipy/scipy/pull/21297/files,

1. Did git-filter-branch for all of scipy/scipy/special
2. Did a scripted interactive rebase, picking only commits which touched a file in one of scipy/scipy/special/special or scipy/scipy/special/xsf.

The interactive rebase was done by setting GIT_SEQUENCE_EDITOR to the script in the details section below. Carrying out the rebase involved a tedious process of manually fixing merge conflicts, typically just using git rm to remove files which were added or updated in scipy/scipy/special but which are not part of the scalar kernel header library. However, sometimes there were merge conflicts because git seems to get confused when you try to rebase through an intermediate merge of main. I made a mistake resolving one of them in xsf/amos.h and needed to fix it in an additional commit here.

I've used git diff --no-index ~/scipy/scipy/special/xsf ~/include/xsf from within the xsf repo to confirm that no unwanted changes have creeped in.

#!/bin/bash

# Get the full commit hashes that touched the directories of interest
COMMITS_TO_KEEP=$(git log --format="%H" -- $HOME/xsf-migration/scipy/include/xsf/scipy/special/special $HOME/xsf-migration/scipy/include/xsf/scipy/special/xsf)
echo $COMMITS_TO_KEEP

# Read the rebase sequence file into an array
mapfile -t LINES < "$1"

# Create a temporary file to store the new rebase sequence
TEMP_FILE=$(mktemp)

# Loop through each line of the rebase sequence
for LINE in "${LINES[@]}"; do
  # Extract the commit hash from the line (this could be variable-length)
    COMMIT_HASH=$(echo "$LINE" | cut -d' ' -f2)

  # Loop through the full commit hashes in COMMITS_TO_KEEP to check for a match
    for FULL_HASH in $COMMITS_TO_KEEP; do
    # Check if the commit hash from the rebase file is a prefix of any full hash
	if [[ "$FULL_HASH" == "$COMMIT_HASH"* ]]; then

      # If there's a match, write the line to the temp file (keep the commit)
      echo "$LINE" >> "$TEMP_FILE"
      break
    fi
  done
done

# Overwrite the original rebase sequence file with the modified one
mv "$TEMP_FILE" "$1"

# Exit successfully
exit 0

[steppi]

Putting this on hold. I talked to @izaid, and there's still a few things that should be done on the SciPy side before we can split things off.

1. Getting MAINT: special: performance optimization for simple autodifferentiation scipy#21568 in.
2. Some updates to API of special functions that can return derivatives of the function. I'll make an RFC about this on scipy soon.
3. Removing dependence of xsf ufunc generation on SciPy's sf_error code.
4. Closing some of the issues from here, META: Streamlined Special Function Development in SciPy SDG Tracking scipy#20223, that @mdhaber and I have been working together on.