Make dump calls conditional to fix errors when building against <2.26

[kounelisagis]

After ce8310e operator<< is used instead of dump functions. Since this is introduced in 2.26.0, building against older libtiledb versions results in failing tests: https://github.com/TileDB-Inc/TileDB-Py/actions/runs/10824513851

Daily tests workflow: https://github.com/TileDB-Inc/TileDB-Py/actions/runs/10832807692

Fixes: #2061

[ihnorton]

Seems correct, but I'm still unclear about what happened here because on latest TileDB-Py+TileDB dev branches, I still get Array type: sparse in the dump output.

In [10]: schema.dump()
- Array type: sparse
- Cell order: row-major
- Tile order: row-major
- Capacity: 10000
- Allows duplicates: false
- Coordinates filters: 1
  > ZSTD: COMPRESSION_LEVEL=-1
- Offsets filters: 1
  > ZSTD: COMPRESSION_LEVEL=-1
- Validity filters: 1
  > RLE: COMPRESSION_LEVEL=-1

### Dimension ###
- Name: d
- Type: STRING_ASCII
- Cell val num: var
- Domain: null
- Tile extent: null
- Filters: 0

### Attribute ###
- Name: a
- Type: STRING_UTF8
- Nullable: true
- Cell val num: var
- Filters: 0
- Fill value:
- Fill value validity: 0
### Current domain ###
- Version: 0
- Empty: 1



In [11]: tiledb.version(), tiledb.libtiledb.version()
Out[11]: ((0, 32, 0), (2, 27, 0))

[kounelisagis]

You are right, dump functions did not change in core at all. (They were only deprecated).

The problem was that operator<< overloads for classes like ArraySchema existed even before TileDB-Inc/TileDB#5026. ArraySchema<SPARSE Domain< Dim<d1,[1,1000],10> Dim<d2,[101,10000],100>> was something expected as an output, no idea why and if it was used by someone.