Update jemallocator. (#3517)

Switch back from tikv-jemallocator to jemallocator as the latter
is now the same as tikv-jemallocator.

The update also brings jemalloc to version 5.3.0, which should
give multiple percent of system level metric improvements.

polars/src/lib.rs

@@ -270,7 +270,7 @@
 //! A DataFrame library naturally does a lot of heap allocations. It is recommended to use a custom
 //! allocator.
 //! [Mimalloc](https://crates.io/crates/mimalloc) and
-//! [JeMalloc](https://crates.io/crates/tikv-jemallocator) for instance, show a significant
+//! [JeMalloc](https://crates.io/crates/jemallocator) for instance, show a significant
 //! performance gain in runtime as well as memory usage.
 //!
 //! #### Usage

py-polars/Cargo.toml

@@ -19,7 +19,7 @@ description = "Blazingly fast DataFrame library"
 mimalloc = { version = "*", default-features = false }
 
 [target.'cfg(target_os="linux")'.dependencies]
-tikv-jemallocator = { version = "0.4", features = ["disable_initial_exec_tls"] }
+jemallocator = { version = "0.5", features = ["disable_initial_exec_tls"] }
 
 [dependencies]
 ahash = "0.7"

py-polars/src/lib.rs

@@ -42,6 +42,8 @@ use crate::error::{
 use crate::file::get_either_file;
 use crate::prelude::{ClosedWindow, DataType, DatetimeArgs, Duration, DurationArgs, PyDataType};
 use dsl::ToExprs;
+#[cfg(target_os = "linux")]
+use jemallocator::Jemalloc;
 #[cfg(not(target_os = "linux"))]
 use mimalloc::MiMalloc;
 use polars::functions::{diag_concat_df, hor_concat_df};
@@ -51,8 +53,6 @@ use polars_core::export::arrow::io::ipc::read::read_file_metadata;
 use polars_core::prelude::IntoSeries;
 use pyo3::panic::PanicException;
 use pyo3::types::{PyBool, PyDict, PyFloat, PyInt, PyString};
-#[cfg(target_os = "linux")]
-use tikv_jemallocator::Jemalloc;
 
 #[global_allocator]
 #[cfg(target_os = "linux")]