Edited read_xdf()

[lgarbar]

The current read_xdf() implementation in this repo fails for some XDF files that I have, likely due to differences in chunk structures or optional timestamps. I’ve made modifications to read_xdf() to:
• Gracefully handle all chunk types without failing on compressed or unusual files.
• Properly pre-allocate and populate arrays for both numeric and string streams.

These changes allow read_xdf() to successfully read all my XDF files while maintaining compatibility with the expected data structures. No other functionality in the repo was altered.

[cbrnr]

Thank you for your contribution! However, I think you are mixing the necessary compatibility fix with several unrelated (and potentially breaking) changes.

The actual fix seems to be string handling for multi-channel data (correct me if I'm wrong, but the related issue should be #10):

if dtype === String
    for ch in 1:nchannels
        nbytes = Int(read_varlen_int(io))
        streams[id]["data"][curr_idx, ch] = String(read(io, nbytes))
    end

The other changes seem unrelated to me, so I suggest reverting them:

1. The function docstring should contain the signature (see https://docs.julialang.org/en/v1/manual/documentation/#Functions-and-Methods), please do not remove it.
2. The Julia logging system already provides functionality for debug and info messages, please do not work against it by introducing a debug parameter (see https://docs.julialang.org/en/v1/stdlib/Logging/).
3. Changing the function signature by (1) introducing a new parameter (as mentioned in the previous point) and (2) changing the sync parameter form positional to keyword-only is a breaking change. Please do not change this unless you think it is useful (but even then I suggest to open a new PR).
4. There are several smaller changes that I'm not sure if they are necessary:
   a. Changing type annotations Dict{Int,Any}() → Dict{Int, Dict}()
   b. Several Int() casts
   c. Renaming variables nsamples → nsamples_block, index[id] → curr_idx
   d. Style changes (2, 3, 4, 6) → (2,3,4,6)
   e. Inconsistent variable naming (both dtype and T is used), and nsamples and index[id] have different names in the second pass (nsamples_block and curr_idx)
   f. The type-specific allocation logic could handle String with the same code as before
   g. No newline at end of file